title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Matched Filters for Noisy Induced Subgraph Detection | The problem of finding the vertex correspondence between two noisy graphs
with different number of vertices where the smaller graph is still large has
many applications in social networks, neuroscience, and computer vision. We
propose a solution to this problem via a graph matching matched filter:
centering and padding the smaller adjacency matrix and applying graph matching
methods to align it to the larger network. The centering and padding schemes
can be incorporated into any algorithm that matches using adjacency matrices.
Under a statistical model for correlated pairs of graphs, which yields a noisy
copy of the small graph within the larger graph, the resulting optimization
problem can be guaranteed to recover the true vertex correspondence between the
networks.
However, there are currently no efficient algorithms for solving this
problem. To illustrate the possibilities and challenges of such problems, we
use an algorithm that can exploit a partially known correspondence and show via
varied simulations and applications to {\it Drosophila} and human connectomes
that this approach can achieve good performance.
| stat |
Projection Pursuit Gaussian Process Regression | A primary goal of computer experiments is to reconstruct the function given
by the computer code via scattered evaluations. Traditional isotropic Gaussian
process models suffer from the curse of dimensionality, when the input
dimension is high. Gaussian process models with additive correlation functions
are scalable to dimensionality, but they are very restrictive as they only work
for additive functions. In this work, we consider a projection pursuit model,
in which the nonparametric part is driven by an additive Gaussian process
regression. The dimension of the additive function is chosen to be higher than
the original input dimension. We show that this dimension expansion can help
approximate more complex functions. A gradient descent algorithm is proposed to
maximize the likelihood function. Simulation studies show that the proposed
method outperforms the traditional Gaussian process models.
| stat |
Models Based on Exponential Interarrival Times for Single-Unusual-Event
Count Data | At least one unusual event appears in some count datasets. It will lead to a
more concentrated (or dispersed) distribution than the Poisson, the gamma, the
Weibull, and the Conway-Maxwell-Poisson (CMP) can accommodate. These well-known
count models are based on the equal rates of interarrival times between
successive events. Under the assumption of unequal rates (one unusual event)
and independent exponential interarrival times, a new class of parametric
models for single-unusual-event (SUE) count data is proposed. These two models
are applied to two empirical applications, the number of births and the number
of bids, and yield considerably better results to the above well-known count
models.
| stat |
Sparse and Smooth Functional Data Clustering | A new model-based procedure is developed for sparse clustering of functional
data that aims to classify a sample of curves into homogeneous groups while
jointly detecting the most informative portions of domain. The proposed method
is referred to as sparse and smooth functional clustering (SaS-Funclust) and
relies on a general functional Gaussian mixture model whose parameters are
estimated by maximizing a log-likelihood function penalized with a functional
adaptive pairwise penalty and a roughness penalty. The former allows
identifying the noninformative portion of domain by shrinking the means of
separated clusters to some common values, whereas the latter improves the
interpretability by imposing some degree of smoothing to the estimated cluster
means. The model is estimated via an expectation-conditional maximization
algorithm paired with a cross-validation procedure. Through a Monte Carlo
simulation study, the SaS-Funclust method is shown to outperform other methods
already appeared in the literature, both in terms of clustering performance and
interpretability. Finally, three real-data examples are presented to
demonstrate the favourable performance of the proposed method. The SaS-Funclust
method is implemented in the $\textsf{R}$ package $\textsf{sasfunclust}$,
available online at https://github.com/unina-sfere/sasfunclust.
| stat |
Analysis of professional basketball field goal attempts via a Bayesian
matrix clustering approach | We propose a Bayesian nonparametric matrix clustering approach to analyze the
latent heterogeneity structure in the shot selection data collected from
professional basketball players in the National Basketball Association (NBA).
The proposed method adopts a mixture of finite mixtures framework and fully
utilizes the spatial information via a mixture of matrix normal distribution
representation. We propose an efficient Markov chain Monte Carlo algorithm for
posterior sampling that allows simultaneous inference on both the number of
clusters and the cluster configurations. We also establish large sample
convergence properties for the posterior distribution. The excellent empirical
performance of the proposed method is demonstrated via simulation studies and
an application to shot chart data from selected players in the 2017 18 NBA
regular season.
| stat |
Propensity Score Weighting for Causal Subgroup Analysis | A common goal in comparative effectiveness research is to estimate treatment
effects on pre-specified subpopulations of patients. Though widely used in
medical research, causal inference methods for such subgroup analysis remain
underdeveloped, particularly in observational studies. In this article, we
develop a suite of analytical methods and visualization tools for causal
subgroup analysis. First, we introduce the estimand of subgroup weighted
average treatment effect and provide the corresponding propensity score
weighting estimator. We show that balancing covariates within a subgroup bounds
the bias of the estimator of subgroup causal effects. Second, we design a new
diagnostic graph -- the Connect-S plot -- for visualizing the subgroup
covariate balance. Finally, we propose to use the overlap weighting method to
achieve exact balance within subgroups. We further propose a method that
combines overlap weighting and LASSO, to balance the bias-variance tradeoff in
subgroup analysis. Extensive simulation studies are presented to compare the
proposed method with several existing methods. We apply the proposed methods to
the Patient-centered Results for Uterine Fibroids (COMPARE-UF) registry data to
evaluate alternative management options for uterine fibroids for relief of
symptoms and quality of life.
| stat |
Investigating the Relationship Between Air Quality and COVID-19
Transmission | It is hypothesized that short-term exposure to air pollution may influence
the transmission of aerosolized pathogens such as COVID-19. We used data from
23 provinces in Italy to build a generalized additive model to investigate the
association between the effective reproductive number of the disease and air
quality while controlling for ambient environmental variables and changes in
human mobility. The model finds that there is a positive, nonlinear
relationship between the density of particulate matter in the air and COVID-19
transmission, which is in alignment with similar studies on other respiratory
illnesses.
| stat |
A Hierarchical Integrative Group LASSO (HiGLASSO) Framework for
Analyzing Environmental Mixtures | Environmental health studies are increasingly measuring multiple pollutants
to characterize the joint health effects attributable to exposure mixtures.
However, the underlying dose-response relationship between toxicants and health
outcomes of interest may be highly nonlinear, with possible nonlinear
interaction effects. Existing penalized regression methods that account for
exposure interactions either cannot accommodate nonlinear interactions while
maintaining strong heredity or are computationally unstable in applications
with limited sample size. In this paper, we propose a general shrinkage and
selection framework to identify noteworthy nonlinear main and interaction
effects among a set of exposures. We design hierarchical integrative group
LASSO (HiGLASSO) to (a) impose strong heredity constraints on two-way
interaction effects (hierarchical), (b) incorporate adaptive weights without
necessitating initial coefficient estimates (integrative), and (c) induce
sparsity for variable selection while respecting group structure (group LASSO).
We prove sparsistency of the proposed method and apply HiGLASSO to an
environmental toxicants dataset from the LIFECODES birth cohort, where the
investigators are interested in understanding the joint effects of 21 urinary
toxicant biomarkers on urinary 8-isoprostane, a measure of oxidative stress. An
implementation of HiGLASSO is available in the higlasso R package, accessible
through the Comprehensive R Archive Network.
| stat |
Targeted VAE: Structured Inference and Targeted Learning for Causal
Parameter Estimation | Undertaking causal inference with observational data is extremely useful
across a wide range of domains including the development of medical treatments,
advertisements and marketing, and policy making. There are two main challenges
associated with undertaking causal inference using observational data:
treatment assignment heterogeneity (i.e., differences between the treated and
untreated groups), and an absence of counterfactual data (i.e. not knowing what
would have happened if an individual who did get treatment, were instead to
have not been treated). We address these two challenges by combining structured
inference and targeted learning. To our knowledge, Targeted Variational
AutoEncoder (TVAE) is the first method to incorporate targeted learning into
deep latent variable models. Results demonstrate competitive and state of the
art performance.
| stat |
Geometric ergodicity of the Random Walk Metropolis with
position-dependent proposal covariance | We consider a Metropolis--Hastings method with proposal $\mathcal{N}(x,
hG(x)^{-1})$, where $x$ is the current state, and study its ergodicity
properties. We show that suitable choices of $G(x)$ can change these compared
to the Random Walk Metropolis case $\mathcal{N}(x, h\Sigma)$, either for better
or worse. We find that if the proposal variance is allowed to grow unboundedly
in the tails of the distribution then geometric ergodicity can be established
when the target distribution for the algorithm has tails that are heavier than
exponential, but that the growth rate must be carefully controlled to prevent
the rejection rate approaching unity. We also illustrate that a judicious
choice of $G(x)$ can result in a geometrically ergodic chain when probability
concentrates on an ever narrower ridge in the tails, something that is not true
for the Random Walk Metropolis.
| stat |
Bayesian Inversion Of Generative Models For Geologic Storage Of Carbon
Dioxide | Carbon capture and storage (CCS) can aid decarbonization of the atmosphere to
limit further global temperature increases. A framework utilizing unsupervised
learning is used to generate a range of subsurface geologic volumes to
investigate potential sites for long-term storage of carbon dioxide. Generative
adversarial networks are used to create geologic volumes, with a further neural
network used to sample the posterior distribution of a trained Generator
conditional to sparsely sampled physical measurements. These generative models
are further conditioned to historic dynamic fluid flow data through Bayesian
inversion to improve the resolution of the forecast of the storage capacity of
injected carbon dioxide.
| stat |
A family of smooth piecewise-linear models with probabilistic
interpretations | The smooth piecewise-linear models cover a wide range of applications
nowadays. Basically, there are two classes of them: models are transitional or
hyperbolic according to their behaviour at the phase-transition zones. This
study explored three different approaches to build smooth piecewise-linear
models, and we analysed their inter-relationships by a unifying modelling
framework. We conceived the smoothed phase-transition zones as domains where a
mixture process takes place, which ensured probabilistic interpretations for
both hyperbolic and transitional models in the light of random thresholds. Many
popular models found in the literature are special cases of our methodology.
Furthermore, this study introduces novel regression models as alternatives,
such as the Epanechnikov, Normal and Skewed-Normal Bent-Cables.
| stat |
A comment on "New non-parametric inferences for low-income proportions"
by Shan Luo and Gengsheng Qin | Shan Luo and Gengsheng Qin published the article "New non-parametric
inferences for low-income proportions" Ann Inst Stat Math, 69, 599-626. In the
note their approach is compared to Zieli\'nski 2009 approach.
| stat |
A Bayesian Nonparametric Approach for Inferring Drug Combination Effects
on Mental Health in People with HIV | Although combination antiretroviral therapy (ART) is highly effective in
suppressing viral load for people with HIV (PWH), many ART agents may
exacerbate central nervous system (CNS)-related adverse effects including
depression. Therefore, understanding the effects of ART drugs on the CNS
function, especially mental health, can help clinicians personalize medicine
with less adverse effects for PWH and prevent them from discontinuing their ART
to avoid undesirable health outcomes and increased likelihood of HIV
transmission. The emergence of electronic health records offers researchers
unprecedented access to HIV data including individuals' mental health records,
drug prescriptions, and clinical information over time. However, modeling such
data is very challenging due to high-dimensionality of the drug combination
space, the individual heterogeneity, and sparseness of the observed drug
combinations. We develop a Bayesian nonparametric approach to learn drug
combination effect on mental health in PWH adjusting for socio-demographic,
behavioral, and clinical factors. The proposed method is built upon the
subset-tree kernel method that represents drug combinations in a way that
synthesizes known regimen structure into a single mathematical representation.
It also utilizes a distance-dependent Chinese restaurant process to cluster
heterogeneous population while taking into account individuals' treatment
histories. We evaluate the proposed approach through simulation studies, and
apply the method to a dataset from the Women's Interagency HIV Study, yielding
interpretable and promising results. Our method has clinical utility in guiding
clinicians to prescribe more informed and effective personalized treatment
based on individuals' treatment histories and clinical characteristics.
| stat |
Learning on heterogeneous graphs using high-order relations | A heterogeneous graph consists of different vertices and edges types.
Learning on heterogeneous graphs typically employs meta-paths to deal with the
heterogeneity by reducing the graph to a homogeneous network, guide random
walks or capture semantics. These methods are however sensitive to the choice
of meta-paths, with suboptimal paths leading to poor performance. In this
paper, we propose an approach for learning on heterogeneous graphs without
using meta-paths. Specifically, we decompose a heterogeneous graph into
different homogeneous relation-type graphs, which are then combined to create
higher-order relation-type representations. These representations preserve the
heterogeneity of edges and retain their edge directions while capturing the
interaction of different vertex types multiple hops apart. This is then
complemented with attention mechanisms to distinguish the importance of the
relation-type based neighbors and the relation-types themselves. Experiments
demonstrate that our model generally outperforms other state-of-the-art
baselines in the vertex classification task on three commonly studied
heterogeneous graph datasets.
| stat |
A Multistep Lyapunov Approach for Finite-Time Analysis of Biased
Stochastic Approximation | Motivated by the widespread use of temporal-difference (TD-) and Q-learning
algorithms in reinforcement learning, this paper studies a class of biased
stochastic approximation (SA) procedures under a mild "ergodic-like" assumption
on the underlying stochastic noise sequence. Building upon a carefully designed
multistep Lyapunov function that looks ahead to several future updates to
accommodate the stochastic perturbations (for control of the gradient bias), we
prove a general result on the convergence of the iterates, and use it to derive
non-asymptotic bounds on the mean-square error in the case of constant
stepsizes. This novel looking-ahead viewpoint renders finite-time analysis of
biased SA algorithms under a large family of stochastic perturbations possible.
For direct comparison with existing contributions, we also demonstrate these
bounds by applying them to TD- and Q-learning with linear function
approximation, under the practical Markov chain observation model. The
resultant finite-time error bound for both the TD- as well as the Q-learning
algorithms is the first of its kind, in the sense that it holds i) for the
unmodified versions (i.e., without making any modifications to the parameter
updates) using even nonlinear function approximators; as well as for Markov
chains ii) under general mixing conditions and iii) starting from any initial
distribution, at least one of which has to be violated for existing results to
be applicable.
| stat |
A Bridge Between Hyperparameter Optimization and Learning-to-learn | We consider a class of a nested optimization problems involving inner and
outer objectives. We observe that by taking into explicit account the
optimization dynamics for the inner objective it is possible to derive a
general framework that unifies gradient-based hyperparameter optimization and
meta-learning (or learning-to-learn). Depending on the specific setting, the
variables of the outer objective take either the meaning of hyperparameters in
a supervised learning problem or parameters of a meta-learner. We show that
some recently proposed methods in the latter setting can be instantiated in our
framework and tackled with the same gradient-based algorithms. Finally, we
discuss possible design patterns for learning-to-learn and present encouraging
preliminary experiments for few-shot learning.
| stat |
Novel Methods for the Analysis of Stepped Wedge Cluster Randomized
Trials | Stepped wedge cluster randomized trials (SW-CRTs) have become increasingly
popular and are used for a variety of interventions and outcomes, often chosen
for their feasibility advantages. SW-CRTs must account for time trends in the
outcome because of the staggered rollout of the intervention inherent in the
design. Robust inference procedures and non-parametric analysis methods have
recently been proposed to handle such trends without requiring strong
parametric modeling assumptions, but these are less powerful than model-based
approaches. We propose several novel analysis methods that reduce reliance on
modeling assumptions while preserving some of the increased power provided by
the use of mixed effects models. In one method, we use the synthetic control
approach to find the best matching clusters for a given intervention cluster.
This approach can improve the power of the analysis but is fully
non-parametric. Another method makes use of within-cluster crossover
information to construct an overall estimator. We also consider methods that
combine these approaches to further improve power. We test these methods on
simulated SW-CRTs and identify settings for which these methods gain robustness
to model misspecification while retaining some of the power advantages of mixed
effects models. Finally, we propose avenues for future research on the use of
these methods; motivation for such research arises from their flexibility,
which allows the identification of specific causal contrasts of interest, their
robustness, and the potential for incorporating covariates to further increase
power. Investigators conducting SW-CRTs might well consider such methods when
common modeling assumptions may not hold.
| stat |
Multilevel models for continuous outcomes | Multilevel models (mixed-effect models or hierarchical linear models) are now
a standard approach to analysing clustered and longitudinal data in the social,
behavioural and medical sciences. This review article focuses on multilevel
linear regression models for continuous responses (outcomes or dependent
variables). These models can be viewed as an extension of conventional linear
regression models to account for and learn from the clustering in the data.
Common clustered applications include studies of school effects on student
achievement, hospital effects on patient health, and neighbourhood effects on
respondent attitudes. In all these examples, multilevel models allow one to
study how the regression relationships vary across clusters, to identify those
cluster characteristics which predict such variation, to disentangle social
processes operating at different levels of analysis, and to make
cluster-specific predictions. Common longitudinal applications include studies
of growth curves of individual height and weight and developmental trajectories
of individual behaviours. In these examples, multilevel models allow one to
describe and explain variation in growth rates and to simultaneously explore
predictors of both of intra- and inter-individual variation. This article
introduces and illustrates this powerful class of model. We start by focusing
on the most commonly applied two-level random-intercept and -slope models. We
illustrate through two detailed examples how these models can be applied to
both clustered and longitudinal data and in both observational and experimental
settings. We then review more flexible three-level, cross-classified, multiple
membership and multivariate response models. We end by recommending a range of
further reading on all these topics.
| stat |
Randomized Graph Cluster Randomization | The global average treatment effect (GATE) is a primary quantity of interest
in the study of causal inference under network interference. With a correctly
specified exposure model of the interference, the Horvitz-Thompson (HT) and
H\'ajek estimators of the GATE are unbiased and consistent, respectively, yet
known to exhibit extreme variance under many designs and in many settings of
interest. With a fixed clustering of the interference graph, graph cluster
randomization (GCR) designs have been shown to greatly reduce variance compared
to node-level random assignment, but even so the variance is still often
prohibitively large. In this work we propose a randomized version of the GCR
design, descriptively named randomized graph cluster randomization (RGCR),
which uses a random clustering rather than a single fixed clustering. By
considering an ensemble of many different cluster assignments, this design
avoids a key problem with GCR where a given node is sometimes "lucky" or
"unlucky" in a given clustering. We propose two randomized graph decomposition
algorithms for use with RGCR, randomized 3-net and 1-hop-max, adapted from
prior work on multiway graph cut problems. When integrating over their own
randomness, these algorithms furnish network exposure probabilities that can be
estimated efficiently. We develop upper bounds on the variance of the HT
estimator of the GATE under assumptions on the metric structure of the
interference graph. Where the best known variance upper bound for the HT
estimator under a GCR design is exponential in the parameters of the metric
structure, we give a comparable variance upper bound under RGCR that is instead
polynomial in the same parameters. We provide extensive simulations comparing
RGCR and GCR designs, observing substantial reductions in the mean squared
error for both HT and H\'ajek estimators of the GATE in a variety of settings.
| stat |
Robust and Reproducible Model Selection Using Bagged Posteriors | Bayesian model selection is premised on the assumption that the data are
generated from one of the postulated models, however, in many applications, all
of these models are incorrect. When two or more models provide a nearly equally
good fit to the data, Bayesian model selection can be highly unstable,
potentially leading to self-contradictory findings. In this paper, we explore
using bagging on the posterior distribution ("BayesBag") when performing model
selection -- that is, averaging the posterior model probabilities over many
bootstrapped datasets. We provide theoretical results characterizing the
asymptotic behavior of the standard posterior and the BayesBag posterior under
misspecification, in the model selection setting. We empirically assess the
BayesBag approach on synthetic and real-world data in (i) feature selection for
linear regression and (ii) phylogenetic tree reconstruction. Our theory and
experiments show that in the presence of misspecification, BayesBag provides
(a) greater reproducibility and (b) greater accuracy in selecting the correct
model, compared to the standard Bayesian posterior; on the other hand, under
correct specification, BayesBag is slightly more conservative than the standard
posterior. Overall, our results demonstrate that BayesBag provides an
easy-to-use and widely applicable approach that improves upon standard Bayesian
model selection by making it more stable and reproducible.
| stat |
Variational Beam Search for Learning with Distribution Shifts | We consider the problem of online learning in the presence of sudden
distribution shifts as frequently encountered in applications such as
autonomous navigation. Distribution shifts require constant performance
monitoring and re-training. They may also be hard to detect and can lead to a
slow but steady degradation in model performance. To address this problem we
propose a new Bayesian meta-algorithm that can both (i) make inferences about
subtle distribution shifts based on minimal sequential observations and (ii)
accordingly adapt a model in an online fashion. The approach uses beam search
over multiple change point hypotheses to perform inference on a hierarchical
sequential latent variable modeling framework. Our proposed approach is
model-agnostic, applicable to both supervised and unsupervised learning, and
yields significant improvements over state-of-the-art Bayesian online learning
approaches.
| stat |
Inference in Stochastic Epidemic Models via Multinomial Approximations | We introduce a new method for inference in stochastic epidemic models which
uses recursive multinomial approximations to integrate over unobserved
variables and thus circumvent likelihood intractability. The method is
applicable to a class of discrete-time, finite-population compartmental models
with partial, randomly under-reported or missing count observations. In
contrast to state-of-the-art alternatives such as Approximate Bayesian
Computation techniques, no forward simulation of the model is required and
there are no tuning parameters. Evaluating the approximate marginal likelihood
of model parameters is achieved through a computationally simple filtering
recursion. The accuracy of the approximation is demonstrated through analysis
of real and simulated data using a model of the 1995 Ebola outbreak in the
Democratic Republic of Congo. We show how the method can be embedded within a
Sequential Monte Carlo approach to estimating the time-varying reproduction
number of COVID-19 in Wuhan, China, recently published by Kucharski et al.
2020.
| stat |
A Note on Bayesian Modeling Specification of Censored Data in JAGS | Just Another Gibbs Sampling (JAGS) is a convenient tool to draw posterior
samples using Markov Chain Monte Carlo for Bayesian modeling. However, the
built-in function dinterval() to model censored data misspecifies the
computation of deviance function, which may limit its usage to perform
likelihood based model comparison. To establish an automatic approach to
specify the correct deviance function in JAGS, we propose a simple alternative
modeling strategy to implement Bayesian model selection for analysis of
censored outcomes. The proposed approach is applicable to a broad spectrum of
data types, which include survival data and many other right-, left- and
interval-censored Bayesian model structures.
| stat |
Detection of the number of principal components by extended AIC-type
method | Estimating the number of principal components is one of the fundamental
problems in many scientific fields such as signal processing (or the spiked
covariance model). In this paper, we first demonstrate that, for fixed $p$, any
penalty term of the form $k'(p-k'/2+1/2)C_n$ may lead to an asymptotically
consistent estimator under the condition that $C_n\to\infty$ and $C_n/n\to0$.
We also extend our results to the case $n,p\to\infty$, with $p/n\to c>0$. In
this case, for $k=o(n^{\frac{1}{3}})$, we first investigate the limiting laws
for the leading eigenvalues of the sample covariance matrix $S_n$ under the
condition that $\lambda_k>1+\sqrt{c}$. At low SNR, since the AIC tends to
underestimate the number of signals $k$, the AIC should be re-defined in this
case. As a natural extension of the AIC for fixed $p$, we propose the extended
AIC (EAIC), i.e., the AIC-type method with tuning parameter
$\gamma=\varphi(c)=1/2+\sqrt{1/c}-\log(1+\sqrt{c})/c$, and demonstrate that the
EAIC-type method, i.e., the AIC-type method with tuning parameter
$\gamma>\varphi(c)$, can select the number of signals $k$ consistently. In the
following two cases, (1) $p$ fixed, $n\to\infty$, (2) $n,p\to\infty$ with
$p/n\to 0$, if the AIC is defined as the degeneration of the EAIC in the case
$n,p\to\infty$ with $p/n\to c>0$, i.e., $\gamma=\lim_{c\rightarrow
0+0}\varphi(c)=1$, then we have essentially demonstrated that, to achieve the
consistency of the AIC-type method in the above two cases, $\gamma>1$ is
required. Moreover, we show that the EAIC-type method is essentially
tuning-free and outperforms the well-known KN estimator proposed in Kritchman
and Nadler (2008) and the BCF estimator proposed in Bai, Choi and Fujikoshi
(2018). Numerical studies indicate that the proposed method works well.
| stat |
GEE-TGDR: A longitudinal feature selection algorithm and its application
to lncRNA expression profiles for psoriasis patients treated with immune
therapies | With the fast evolution of high-throughput technology, longitudinal gene
expression experiments have become affordable and increasingly common in
biomedical fields. Generalized estimating equation (GEE) approach is a widely
used statistical method for the analysis of longitudinal data. Feature
selection is imperative in longitudinal omics data analysis. Among a variety of
existing feature selection methods, an embedded method, namely, threshold
gradient descent regularization (TGDR) stands out due to its excellent
characteristics. An alignment of GEE with TGDR is a promising area for the
purpose of identifying relevant markers that can explain the dynamic changes of
outcomes across time. In this study, we proposed a new novel feature selection
algorithm for longitudinal outcomes:GEE-TGDR. In the GEE-TGDR method, the
corresponding quasi-likelihood function of a GEE model is the objective
function to be optimized and the optimization and feature selection are
accomplished by the TGDR method. We applied the GEE-TGDR method a longitudinal
lncRNA gene expression dataset that examined the treatment response of
psoriasis patients to immune therapy. Under different working correlation
structures, a list including 10 relevant lncRNAs were identified with a
predictive accuracy of 80 % and meaningful biological interpretation. To
conclude, a widespread application of the proposed GEE-TGDR method in omics
data analysis is anticipated.
| stat |
Quantile-based clustering | A new cluster analysis method, $K$-quantiles clustering, is introduced.
$K$-quantiles clustering can be computed by a simple greedy algorithm in the
style of the classical Lloyd's algorithm for $K$-means. It can be applied to
large and high-dimensional datasets. It allows for within-cluster skewness and
internal variable scaling based on within-cluster variation. Different versions
allow for different levels of parsimony and computational efficiency. Although
$K$-quantiles clustering is conceived as nonparametric, it can be connected to
a fixed partition model of generalized asymmetric Laplace-distributions. The
consistency of $K$-quantiles clustering is proved, and it is shown that
$K$-quantiles clusters correspond to well separated mixture components in a
nonparametric mixture. In a simulation, $K$-quantiles clustering is compared
with a number of popular clustering methods with good results. A
high-dimensional microarray dataset is clustered by $K$-quantiles.
| stat |
Decline of COPD exacerbations in clinical trials over two decades -- a
systematic review and meta-regression | BACKGROUND: An important goal of chronic obstructive pulmonary disease (COPD)
treatment is to reduce the frequency of exacerbations. Some observations
suggest a decline in exacerbation rates in clinical trials over time. A more
systematic understanding would help to improve the design and interpretation of
COPD trials.
METHODS: We performed a systematic review and meta-regression of the placebo
groups in published randomized controlled trials reporting exacerbations as an
outcome. A Bayesian negative binomial model was developed to accommodate
results that are reported in different formats; results are reported with
credible intervals (CI) and posterior tail probabilities ($p_B$).
RESULTS: Of 1114 studies identified by our search, 55 were ultimately
included. Exacerbation rates decreased by 6.7% (95% CI (4.4, 9.0); $p_B$ <
0.001) per year, or 50% (95% CI (36, 61)) per decade. Adjusting for available
study and baseline characteristics such as forced expiratory volume in 1 s
(FEV1) did not alter the observed trend considerably. Two subsets of studies,
one using a true placebo group and the other allowing inhaled corticosteroids
in the "placebo" group, also yielded consistent results.
CONCLUSIONS: In conclusion, this meta-regression indicates that the rate of
COPD exacerbations decreased over the past two decades to a clinically relevant
extent independent of important prognostic factors. This suggests that care is
needed in the design of new trials or when comparing results from older trials
with more recent ones. Also a considerable effect of adjunct therapy on COPD
exacerbations can be assumed.
| stat |
The population-attributable fraction for time-dependent exposures using
dynamic prediction and landmarking | The public health impact of a harmful exposure can be quantified by the
population-attributable fraction (PAF). The PAF describes the attributable risk
due to an exposure and is often interpreted as the proportion of preventable
cases if the exposure could be extinct. Difficulties in the definition and
interpretation of the PAF arise when the exposure of interest depends on time.
Then, the definition of exposed and unexposed individuals is not
straightforward. We propose dynamic prediction and landmarking to define and
estimate a PAF in this data situation. Two estimands are discussed which are
based on two hypothetical interventions that could prevent the exposure in
different ways. Considering the first estimand, at each landmark the estimation
problem is reduced to a time-independent setting. Then, estimation is simply
performed by using a generalized-linear model accounting for the current
exposure state and further (time-varying) covariates. The second estimand is
based on counterfactual outcomes, estimation can be performed using
pseudo-values or inverse-probability weights. The approach is explored in a
simulation study and applied on two data examples. First, we study a large
French database of intensive care unit patients to estimate the
population-benefit of a pathogen-specific intervention that could prevent
ventilator-associated pneumonia caused by the pathogen Pseudomonas aeruginosa.
Moreover, we quantify the population-attributable burden of locoregional and
distant recurrence in breast cancer patients.
| stat |
Switching Linear Dynamics for Variational Bayes Filtering | System identification of complex and nonlinear systems is a central problem
for model predictive control and model-based reinforcement learning. Despite
their complexity, such systems can often be approximated well by a set of
linear dynamical systems if broken into appropriate subsequences. This
mechanism not only helps us find good approximations of dynamics, but also
gives us deeper insight into the underlying system. Leveraging Bayesian
inference, Variational Autoencoders and Concrete relaxations, we show how to
learn a richer and more meaningful state space, e.g. encoding joint constraints
and collisions with walls in a maze, from partial and high-dimensional
observations. This representation translates into a gain of accuracy of learned
dynamics showcased on various simulated tasks.
| stat |
Quickest detection in practice in presence of seasonality: An
illustration with call center data | In this chapter, we explain how quickest detection algorithms can be useful
for risk management in presence of seasonality. We investigate the problem of
detecting fast enough cases when a call center will need extra staff in a near
future with a high probability. We illustrate our findings on real data
provided by a French insurer. We also discuss the relevance of the CUSUM
algorithm and of some machine-learning type competitor for this applied
problem.
| stat |
Quantifying Uncertainty with a Derivative Tracking SDE Model and
Application to Wind Power Forecast Data | We develop a data-driven methodology based on parametric It\^{o}'s Stochastic
Differential Equations (SDEs) to capture the real asymmetric dynamics of
forecast errors. Our SDE framework features time-derivative tracking of the
forecast, time-varying mean-reversion parameter, and an improved
state-dependent diffusion term. Proofs of the existence, strong uniqueness, and
boundedness of the SDE solutions are shown under a principled condition for the
time-varying mean-reversion parameter. Inference based on approximate
likelihood, constructed through the moment-matching technique both in the
original forecast error space and in the Lamperti space, is performed through
numerical optimization procedures. We propose another contribution based on the
fixed-point likelihood optimization approach in the Lamperti space. All the
procedures are agnostic of the forecasting technology, and they enable
comparisons between different forecast providers. We apply our SDE framework to
model historical Uruguayan normalized wind power production and forecast data
between April and December 2019. Sharp empirical confidence bands of future
wind power production are obtained for the best-selected model.
| stat |
Optimized data exploration applied to the simulation of a chemical
process | In complex simulation environments, certain parameter space regions may
result in non-convergent or unphysical outcomes. All parameters can therefore
be labeled with a binary class describing whether or not they lead to valid
results. In general, it can be very difficult to determine feasible parameter
regions, especially without previous knowledge. We propose a novel algorithm to
explore such an unknown parameter space and improve its feasibility
classification in an iterative way. Moreover, we include an additional
optimization target in the algorithm to guide the exploration towards regions
of interest and to improve the classification therein. In our method we make
use of well-established concepts from the field of machine learning like kernel
support vector machines and kernel ridge regression. From a comparison with a
Kriging-based exploration approach based on recently published results we can
show the advantages of our algorithm in a binary feasibility classification
scenario with a discrete feasibility constraint violation. In this context, we
also propose an improvement of the Kriging-based exploration approach. We apply
our novel method to a fully realistic, industrially relevant chemical process
simulation to demonstrate its practical usability and find a comparably good
approximation of the data space topology from relatively few data points.
| stat |
Global Sensitivity and Domain-Selective Testing for Functional-Valued
Responses: An Application to Climate Economy Models | Complex computational models are increasingly used by business and
governments for making decisions, such as how and where to invest to transition
to a low carbon world. Complexity arises with great evidence in the outputs
generated by large scale models, and calls for the use of advanced Sensitivity
Analysis techniques. To our knowledge, there are no methods able to perform
sensitivity analysis for outputs that are more complex than scalar ones and to
deal with model uncertainty using a sound statistical framework. The aim of
this work is to address these two shortcomings by combining sensitivity and
functional data analysis. We express output variables as smooth functions,
employing a Functional Data Analysis (FDA) framework. We extend global
sensitivity techniques to function-valued responses and perform significance
testing over sensitivity indices. We apply the proposed methods to computer
models used in climate economics. While confirming the qualitative intuitions
of previous works, we are able to test the significance of input assumptions
and of their interactions. Moreover, the proposed method allows to identify the
time dynamics of sensitivity indices.
| stat |
Developing Biomarker Combinations in Multicenter Studies via Direct
Maximization and Penalization | Motivated by a study of acute kidney injury, we consider the setting of
biomarker studies involving patients at multiple centers where the goal is to
develop a biomarker combination for diagnosis, prognosis, or screening. As
biomarker studies become larger, this type of data structure will be
encountered more frequently. In the presence of multiple centers, one way to
assess the predictive capacity of a given combination is to consider the
center-adjusted AUC (aAUC), a summary of the ability of the combination to
discriminate between cases and controls in each center. Rather than using a
general method, such as logistic regression, to construct the biomarker
combination, we propose directly maximizing the aAUC. Furthermore, it may be
desirable to have a biomarker combination with similar performance across
centers. To that end, we allow for penalization of the variability in the
center-specific AUCs. We demonstrate desirable asymptotic properties of the
resulting combinations. Simulations provide small-sample evidence that
maximizing the aAUC can lead to combinations with improved performance. We also
use simulated data to illustrate the utility of constructing combinations by
maximizing the aAUC while penalizing variability. Finally, we apply these
methods to data from the study of acute kidney injury.
| stat |
Using prior expansions for prior-data conflict checking | Any Bayesian analysis involves combining information represented through
different model components, and when different sources of information are in
conflict it is important to detect this. Here we consider checking for
prior-data conflict in Bayesian models by expanding the prior used for the
analysis into a larger family of priors, and considering a marginal likelihood
score statistic for the expansion parameter. Consideration of different
expansions can be informative about the nature of any conflict, and extensions
to hierarchically specified priors and connections with other approaches to
prior-data conflict checking are discussed. Implementation in complex
situations is illustrated with two applications. The first concerns testing for
the appropriateness of a LASSO penalty in shrinkage estimation of coefficients
in linear regression. Our method is compared with a recent suggestion in the
literature designed to be powerful against alternatives in the exponential
power family, and we use this family as the prior expansion for constructing
our check. A second application concerns a problem in quantum state estimation,
where a multinomial model is considered with physical constraints on the model
parameters. In this example, the usefulness of different prior expansions is
demonstrated for obtaining checks which are sensitive to different aspects of
the prior.
| stat |
Additive quantile regression for clustered data with an application to
children's physical activity | Additive models are flexible regression tools that handle linear as well as
nonlinear terms. The latter are typically modelled via smoothing splines.
Additive mixed models extend additive models to include random terms when the
data are sampled according to cluster designs (e.g., longitudinal). These
models find applications in the study of phenomena like growth, certain disease
mechanisms and energy consumption in humans, when repeated measurements are
available. In this paper, we propose a novel additive mixed model for quantile
regression. Our methods are motivated by an application to physical activity
based on a dataset with more than half million accelerometer measurements in
children of the UK Millennium Cohort Study. In a simulation study, we assess
the proposed methods against existing alternatives.
| stat |
Identification of unknown parameters and prediction with hierarchical
matrices | Statistical analysis of massive datasets very often implies expensive linear
algebra operations with large dense matrices. Typical tasks are an estimation
of unknown parameters of the underlying statistical model and prediction of
missing values. We developed the H-MLE procedure, which solves these tasks. The
unknown parameters can be estimated by maximizing the joint Gaussian
log-likelihood function, which depends on a covariance matrix. To decrease high
computational cost, we approximate the covariance matrix in the hierarchical
(H-) matrix format. The H-matrix technique allows us to work with inhomogeneous
covariance matrices and almost arbitrary locations. Especially, H-matrices can
be applied in cases when the matrices under consideration are dense and
unstructured.
For validation purposes, we implemented three machine learning methods: the
k-nearest neighbors (kNN), random forest, and deep neural network. The best
results (for the given datasets) were obtained by the kNN method with three or
seven neighbors depending on the dataset. The results computed with the H-MLE
method were compared with the results obtained by the kNN method.
The developed H-matrix code and all datasets are freely available online.
| stat |
Approximate Inference Turns Deep Networks into Gaussian Processes | Deep neural networks (DNN) and Gaussian processes (GP) are two powerful
models with several theoretical connections relating them, but the relationship
between their training methods is not well understood. In this paper, we show
that certain Gaussian posterior approximations for Bayesian DNNs are equivalent
to GP posteriors. This enables us to relate solutions and iterations of a
deep-learning algorithm to GP inference. As a result, we can obtain a GP kernel
and a nonlinear feature map while training a DNN. Surprisingly, the resulting
kernel is the neural tangent kernel. We show kernels obtained on real datasets
and demonstrate the use of the GP marginal likelihood to tune hyperparameters
of DNNs. Our work aims to facilitate further research on combining DNNs and GPs
in practical settings.
| stat |
Rank-deficiencies in a reduced information latent variable model | Latent variable models are well-known to suffer from rank deficiencies,
causing problems with convergence and stability. Such problems are compounded
in the "reduced-group split-ballot multitrait-multimethod model", which omits a
set of moments from the estimation through a planned missing data design. This
paper demonstrates the existence of rank deficiencies in this model and give
the explicit null space. It also demonstrates that sample size and distance
from the rank-deficient point interact in their effects on convergence, causing
convergence to improve or worsen depending on both factors simultaneously.
Furthermore, it notes that the latent variable correlations in the uncorrelated
methods SB-MTMM model remain unaffected by the rank deficiency. I conclude that
methodological experiments should be careful to manipulate both distance to
known rank-deficiencies and sample size, and report all results, not only the
apparently converged ones. Practitioners may consider that, even in the
presence of nonconvergence or so-called "inadmissible" estimates, a subset of
parameter estimates may still be consistent and stable.
| stat |
Scalable Thompson Sampling via Optimal Transport | Thompson sampling (TS) is a class of algorithms for sequential
decision-making, which requires maintaining a posterior distribution over a
model. However, calculating exact posterior distributions is intractable for
all but the simplest models. Consequently, efficient computation of an
approximate posterior distribution is a crucial problem for scalable TS with
complex models, such as neural networks. In this paper, we use distribution
optimization techniques to approximate the posterior distribution, solved via
Wasserstein gradient flows. Based on the framework, a principled
particle-optimization algorithm is developed for TS to approximate the
posterior efficiently. Our approach is scalable and does not make explicit
distribution assumptions on posterior approximations. Extensive experiments on
both synthetic data and real large-scale data demonstrate the superior
performance of the proposed methods.
| stat |
Introduction to Dynamic Linear Models for Time Series Analysis | Dynamic linear models (DLM) offer a very generic framework to analyse time
series data. Many classical time series models can be formulated as DLMs,
including ARMA models and standard multiple linear regression models. The
models can be seen as general regression models where the coefficients can vary
in time. In addition, they allow for a state space representation and a
formulation as hierarchical statistical models, which in turn is the key for
efficient estimation by Kalman formulas and by Markov chain Monte Carlo (MCMC)
methods. A dynamic linear model can handle non-stationary processes, missing
values and non-uniform sampling as well as observations with varying
accuracies. This chapter gives an introduction to DLM and shows how to build
various useful models for analysing trends and other sources of variability in
geodetic time series.
| stat |
Functional Group Bridge for Simultaneous Regression and Support
Estimation | This article is motivated by studying multisensory effects on brain
activities in intracranial electroencephalography (iEEG) experiments.
Differential brain activities to multisensory stimulus presentations are zero
in most regions and non-zero in some local regions, yielding locally sparse
functions. Such studies are essentially a function-on-scalar regression
problem, with interest being focused not only on estimating nonparametric
functions but also on recovering the function supports. We propose a weighted
group bridge approach for simultaneous function estimation and support recovery
in function-on-scalar mixed effect models, while accounting for heterogeneity
present in functional data. We use B-splines to transform sparsity of functions
to its sparse vector counterpart of increasing dimension, and propose a fast
non-convex optimization algorithm using nested alternative direction method of
multipliers (ADMM) for estimation. Large sample properties are established. In
particular, we show that the estimated coefficient functions are rate optimal
in the minimax sense under the $L_2$ norm and resemble a phase transition
phenomenon. For support estimation, we derive a convergence rate under the
$L_{\infty}$ norm that leads to a sparsistency property under
$\delta$-sparsity, and provide a simple sufficient regularity condition under
which a strict sparsistency property is established. An adjusted extended
Bayesian information criterion is proposed for parameter tuning. The developed
method is illustrated through simulation and an application to a novel iEEG
dataset to study multisensory integration. We integrate the proposed method
into RAVE, an R package that gains increasing popularity in the iEEG community.
| stat |
Probabilistic Time Series Forecasting with Structured Shape and Temporal
Diversity | Probabilistic forecasting consists in predicting a distribution of possible
future outcomes. In this paper, we address this problem for non-stationary time
series, which is very challenging yet crucially important. We introduce the
STRIPE model for representing structured diversity based on shape and time
features, ensuring both probable predictions while being sharp and accurate.
STRIPE is agnostic to the forecasting model, and we equip it with a
diversification mechanism relying on determinantal point processes (DPP). We
introduce two DPP kernels for modeling diverse trajectories in terms of shape
and time, which are both differentiable and proved to be positive
semi-definite. To have an explicit control on the diversity structure, we also
design an iterative sampling mechanism to disentangle shape and time
representations in the latent space. Experiments carried out on synthetic
datasets show that STRIPE significantly outperforms baseline methods for
representing diversity, while maintaining accuracy of the forecasting model. We
also highlight the relevance of the iterative sampling scheme and the
importance to use different criteria for measuring quality and diversity.
Finally, experiments on real datasets illustrate that STRIPE is able to
outperform state-of-the-art probabilistic forecasting approaches in the best
sample prediction.
| stat |
Reply to "Issues arising from benchmarking single-cell RNA sequencing
imputation methods" | In our Brief Communication (DOI: 10.1038/s41592-018-0033-z), we presented the
method SAVER for recovering true gene expression levels in noisy single cell
RNA sequencing data. We evaluated the performance of SAVER, along with
comparable methods MAGIC and scImpute, in an RNA FISH validation experiment and
a data downsampling experiment. In a Comment [arXiv:1908.07084v1], Li & Li were
concerned with the use of the downsampled datasets, specifically focusing on
clustering results obtained from the Zeisel et al. data. Here, we will address
these comments and, furthermore, amend the data downsampling experiment to
demonstrate that the findings from the data downsampling experiment in our
Brief Communication are valid.
| stat |
On Riemannian Stochastic Approximation Schemes with Fixed Step-Size | This paper studies fixed step-size stochastic approximation (SA) schemes,
including stochastic gradient schemes, in a Riemannian framework. It is
motivated by several applications, where geodesics can be computed explicitly,
and their use accelerates crude Euclidean methods. A fixed step-size scheme
defines a family of time-homogeneous Markov chains, parametrized by the
step-size. Here, using this formulation, non-asymptotic performance bounds are
derived, under Lyapunov conditions. Then, for any step-size, the corresponding
Markov chain is proved to admit a unique stationary distribution, and to be
geometrically ergodic. This result gives rise to a family of stationary
distributions indexed by the step-size, which is further shown to converge to a
Dirac measure, concentrated at the solution of the problem at hand, as the
step-size goes to 0. Finally, the asymptotic rate of this convergence is
established, through an asymptotic expansion of the bias, and a central limit
theorem.
| stat |
Composite likelihood methods for histogram-valued random variables | Symbolic data analysis has been proposed as a technique for summarising large
and complex datasets into a much smaller and tractable number of distributions
-- such as random rectangles or histograms -- each describing a portion of the
larger dataset. Recent work has developed likelihood-based methods that permit
fitting models for the underlying data while only observing the distributional
summaries. However, while powerful, when working with random histograms this
approach rapidly becomes computationally intractable as the dimension of the
underlying data increases. We introduce a composite-likelihood variation of
this likelihood-based approach for the analysis of random histograms in $K$
dimensions, through the construction of lower-dimensional marginal histograms.
The performance of this approach is examined through simulated and real data
analysis of max-stable models for spatial extremes using millions of observed
datapoints in more than $K=100$ dimensions. Large computational savings are
available compared to existing model fitting approaches.
| stat |
Uncertainty Autoencoders: Learning Compressed Representations via
Variational Information Maximization | Compressed sensing techniques enable efficient acquisition and recovery of
sparse, high-dimensional data signals via low-dimensional projections. In this
work, we propose Uncertainty Autoencoders, a learning framework for
unsupervised representation learning inspired by compressed sensing. We treat
the low-dimensional projections as noisy latent representations of an
autoencoder and directly learn both the acquisition (i.e., encoding) and
amortized recovery (i.e., decoding) procedures. Our learning objective
optimizes for a tractable variational lower bound to the mutual information
between the datapoints and the latent representations. We show how our
framework provides a unified treatment to several lines of research in
dimensionality reduction, compressed sensing, and generative modeling.
Empirically, we demonstrate a 32% improvement on average over competing
approaches for the task of statistical compressed sensing of high-dimensional
datasets.
| stat |
Probabilistic reconstruction of truncated particle trajectories on a
closed surface | Investigation of dynamic processes in cell biology very often relies on the
observation in two dimensions of 3D biological processes. Consequently, the
data are partial and statistical methods and models are required to recover the
parameters describing the dynamical processes. In the case of molecules moving
over the 3D surface, such as proteins on walls of bacteria cell, a large
portion of the 3D surface is not observed in 2D-time microscopy. It follows
that biomolecules may disappear for a period of time in a region of interest,
and then reappear later. Assuming Brownian motion with drift, we address the
mathematical problem of the reconstruction of biomolecules trajectories on a
cylindrical surface. A subregion of the cylinder is typically recorded during
the observation period, and biomolecules may appear or disappear in any place
of the 3D surface. The performance of the method is demonstrated on simulated
particle trajectories that mimic MreB protein dynamics observed in 2D
time-lapse fluorescence microscopy in rod-shaped bacteria.
| stat |
Estimating spillovers using imprecisely measured networks | In many experimental contexts, whether and how network interactions impact
the outcome of interest for both treated and untreated individuals are key
concerns. Networks data is often assumed to perfectly represent these possible
interactions. This paper considers the problem of estimating treatment effects
when measured connections are, instead, a noisy representation of the true
spillover pathways. We show that existing methods, using the potential outcomes
framework, yield biased estimators in the presence of this mismeasurement. We
develop a new method, using a class of mixture models, that can account for
missing connections and discuss its estimation via the Expectation-Maximization
algorithm. We check our method's performance by simulating experiments on real
network data from 43 villages in India. Finally, we use data from a previously
published study to show that estimates using our method are more robust to the
choice of network measure.
| stat |
Model combinations through revised base-rates | Standard selection criteria for forecasting models focus on information that
is calculated for each series independently, disregarding the general
tendencies and performances of the candidate models. In this paper, we propose
a new way to statistical model selection and model combination that
incorporates the base-rates of the candidate forecasting models, which are then
revised so that the per-series information is taken into account. We examine
two schemes that are based on the precision and sensitivity information from
the contingency table of the base rates. We apply our approach on pools of
exponential smoothing models and a large number of real time series and we show
that our schemes work better than standard statistical benchmarks. We discuss
the connection of our approach to other cross-learning approaches and offer
insights regarding implications for theory and practice.
| stat |
Cluster analysis and outlier detection with missing data | A mixture of multivariate contaminated normal (MCN) distributions is a useful
model-based clustering technique to accommodate data sets with mild outliers.
However, this model only works when fitted to complete data sets, which is
often not the case in real applications. In this paper, we develop a framework
for fitting a mixture of MCN distributions to incomplete data sets, i.e. data
sets with some values missing at random. We employ the expectation-conditional
maximization algorithm for parameter estimation. We use a simulation study to
compare the results of our model and a mixture of Student's t distributions for
incomplete data.
| stat |
Combining interdependent climate model outputs in CMIP5: A spatial
Bayesian approach | Projections of future climate change rely heavily on climate models, and
combining climate models through a multi-model ensemble is both more accurate
than a single climate model and valuable for uncertainty quantification.
However, Bayesian approaches to multi-model ensembles have been criticized for
making oversimplified assumptions about bias and variability, as well as
treating different models as statistically independent. This paper extends the
Bayesian hierarchical approach of Sansom et al. (2017) by explicitly accounting
for spatial variability and inter-model dependence. We propose a Bayesian
hierarchical model that accounts for bias between climate models and
observations, spatial and inter-model dependence, the emergent relationship
between historical and future periods, and natural variability. Extensive
simulations show that our model provides better estimates and uncertainty
quantification than the commonly used simple model mean. These results are
illustrated using data from the CMIP5 model archive. As examples, for Central
North America our projected mean temperature for 2070--2100 is about 0.8 K
lower than the simple model mean, while for East Asia it is about 0.5 K higher;
however, in both cases, the widths of the 90% credible intervals are of the
order 3--6 K, so the uncertainties overwhelm the relatively small differences
in projected mean temperatures.
| stat |
A Proposed Method for Assessing Cluster Heterogeneity | Assessing how adequate clusters fit a dataset and finding an optimum number
of clusters is a difficult process. A membership matrix and the degree of
membership matrix is suggested to determine the homogeneity of a cluster fit.
Maximisation of the ratio of the overall degree of membership at cluster number
lag 1 is also suggested as a method to optimise the number of clusters in a
dataset. A threshold factor upon the degree of membership is also suggested for
homogeneous clusters. Cluster simulations were given to compare how well the
proposed method compares against established methods. This method may be
applied to the output of both hierarchical and k-means clustering.
| stat |
Non-linear Mediation Analysis with High-dimensional Mediators whose
Causal Structure is Unknown | With multiple potential mediators on the causal pathway from a treatment to
an outcome, we consider the problem of decomposing the effects along multiple
possible causal path(s) through each distinct mediator. Under Pearl's
path-specific effects framework (Pearl, 2001; Avin et al., 2005), such
fine-grained decompositions necessitate stringent assumptions, such as
correctly specifying the causal structure among the mediators, and there being
no unobserved confounding among the mediators. In contrast, interventional
direct and indirect effects for multiple mediators (Vansteelandt and Daniel,
2017) can be identified under much weaker conditions, while providing
scientifically relevant causal interpretations. Nonetheless, current estimation
approaches require (correctly) specifying a model for the joint mediator
distribution, which can be difficult when there is a high-dimensional set of
possibly continuous and non-continuous mediators. In this article, we avoid the
need to model this distribution, by developing a definition of interventional
effects previously suggested by VanderWeele and Tchetgen Tchetgen (2017) for
longitudinal mediation. We propose a novel estimation strategy that uses
non-parametric estimates of the (counterfactual) mediator distributions.
Non-continuous outcomes can be accommodated using non-linear outcome models.
Estimation proceeds via Monte Carlo integration. The procedure is illustrated
using publicly available genomic data (Huang and Pan, 2016) to assess the
causal effect of a microRNA expression on the three-month mortality of brain
cancer patients that is potentially mediated by expression values of multiple
genes.
| stat |
Efficient Optimistic Exploration in Linear-Quadratic Regulators via
Lagrangian Relaxation | We study the exploration-exploitation dilemma in the linear quadratic
regulator (LQR) setting. Inspired by the extended value iteration algorithm
used in optimistic algorithms for finite MDPs, we propose to relax the
optimistic optimization of \ofulq and cast it into a constrained
\textit{extended} LQR problem, where an additional control variable implicitly
selects the system dynamics within a confidence interval. We then move to the
corresponding Lagrangian formulation for which we prove strong duality. As a
result, we show that an $\epsilon$-optimistic controller can be computed
efficiently by solving at most $O\big(\log(1/\epsilon)\big)$ Riccati equations.
Finally, we prove that relaxing the original \ofu problem does not impact the
learning performance, thus recovering the $\tilde{O}(\sqrt{T})$ regret of
\ofulq. To the best of our knowledge, this is the first computationally
efficient confidence-based algorithm for LQR with worst-case optimal regret
guarantees.
| stat |
The Boomerang Sampler | This paper introduces the Boomerang Sampler as a novel class of
continuous-time non-reversible Markov chain Monte Carlo algorithms. The
methodology begins by representing the target density as a density, $e^{-U}$,
with respect to a prescribed (usually) Gaussian measure and constructs a
continuous trajectory consisting of a piecewise elliptical path. The method
moves from one elliptical orbit to another according to a rate function which
can be written in terms of $U$. We demonstrate that the method is easy to
implement and demonstrate empirically that it can out-perform existing
benchmark piecewise deterministic Markov processes such as the bouncy particle
sampler and the Zig-Zag. In the Bayesian statistics context, these competitor
algorithms are of substantial interest in the large data context due to the
fact that they can adopt data subsampling techniques which are exact (ie induce
no error in the stationary distribution). We demonstrate theoretically and
empirically that we can also construct a control-variate subsampling boomerang
sampler which is also exact, and which possesses remarkable scaling properties
in the large data limit. We furthermore illustrate a factorised version on the
simulation of diffusion bridges.
| stat |
Functional Models for Time-Varying Random Objects | In recent years, samples of time-varying object data such as time-varying
networks that are not in a vector space have been increasingly collected. These
data can be viewed as elements of a general metric space that lacks local or
global linear structure and therefore common approaches that have been used
with great success for the analysis of functional data, such as functional
principal component analysis, cannot be applied. In this paper we propose
metric covariance, a novel association measure for paired object data lying in
a metric space $(\Omega,d)$ that we use to define a metric auto-covariance
function for a sample of random $\Omega$-valued curves, where $\Omega$
generally will not have a vector space or manifold structure. The proposed
metric auto-covariance function is non-negative definite when the squared
semimetric $d^2$ is of negative type. Then the eigenfunctions of the linear
operator with the auto-covariance function as kernel can be used as building
blocks for an object functional principal component analysis for
$\Omega$-valued functional data, including time-varying probability
distributions, covariance matrices and time-dynamic networks. Analogues of
functional principal components for time-varying objects are obtained by
applying Fr\'echet means and projections of distance functions of the random
object trajectories in the directions of the eigenfunctions, leading to
real-valued Fr\'echet scores. Using the notion of generalized Fr\'echet
integrals, we construct object functional principal components that lie in the
metric space $\Omega$.
| stat |
On Stationary-Point Hitting Time and Ergodicity of Stochastic Gradient
Langevin Dynamics | Stochastic gradient Langevin dynamics (SGLD) is a fundamental algorithm in
stochastic optimization. Recent work by Zhang et al. [2017] presents an
analysis for the hitting time of SGLD for the first and second order stationary
points. The proof in Zhang et al. [2017] is a two-stage procedure through
bounding the Cheeger's constant, which is rather complicated and leads to loose
bounds. In this paper, using intuitions from stochastic differential equations,
we provide a direct analysis for the hitting times of SGLD to the first and
second order stationary points. Our analysis is straightforward. It only relies
on basic linear algebra and probability theory tools. Our direct analysis also
leads to tighter bounds comparing to Zhang et al. [2017] and shows the explicit
dependence of the hitting time on different factors, including dimensionality,
smoothness, noise strength, and step size effects. Under suitable conditions,
we show that the hitting time of SGLD to first-order stationary points can be
dimension-independent. Moreover, we apply our analysis to study several
important online estimation problems in machine learning, including linear
regression, matrix factorization, and online PCA.
| stat |
Robust penalized empirical likelihood in high dimensional longitudinal
data analysis | As an effective nonparametric method, empirical likelihood (EL) is appealing
in combining estimating equations flexibly and adaptively for incorporating
data information. To select important variables in the sparse high-dimensional
model, we consider a penalized EL method based on robust estimating functions
by applying two penalty functions for regularizing the regression parameters
and the associated Lagrange multipliers simultaneously, which allows the
dimensionalities of both regression parameters and estimating equations to grow
exponentially with the sample size. The proposed method can improve the
robustness and effectiveness when the data have underlying outliers or heavy
tails in the response variables and/or covariates. The oracle properties are
also established under some regularity conditions. Extensive simulation studies
and a yeast cell data are used to evaluate the performance of the proposed
method. The numerical results reveal that robust remedies on estimating
equations are needed when the data have heavy tails and/or include underlying
outliers.
| stat |
Graph-aware Modeling of Brain Connectivity Networks | Functional connections in the brain are frequently represented by weighted
networks, with nodes representing locations in the brain, and edges
representing the strength of connectivity between these locations. One
challenge in analyzing such data is that inference at the individual edge level
is not particularly biologically meaningful; interpretation is more useful at
the level of so-called functional regions, or groups of nodes and connections
between them; this is often called "graph-aware" inference in the neuroimaging
literature. However, pooling over functional regions leads to significant loss
of information and lower accuracy. Another challenge is correlation among edge
weights within a subject, which makes inference based on independence
assumptions unreliable. We address both these challenges with a linear mixed
effects model, which accounts for functional regions and for edge dependence,
while still modeling individual edge weights to avoid loss of information. The
model allows for comparing two populations, such as patients and healthy
controls, both at the functional regions level and at individual edge level,
leading to biologically meaningful interpretations. We fit this model to a
resting state fMRI data on schizophrenics and healthy controls, obtaining
interpretable results consistent with the schizophrenia literature.
| stat |
Function approximation by deep neural networks with parameters $\{0,\pm
\frac{1}{2}, \pm 1, 2\}$ | In this paper it is shown that $C_\beta$-smooth functions can be approximated
by neural networks with parameters $\{0,\pm \frac{1}{2}, \pm 1, 2\}$. The
depth, width and the number of active parameters of constructed networks have,
up to a logarithimc factor, the same dependence on the approximation error as
the networks with parameters in $[-1,1]$. In particular, this means that the
nonparametric regression estimation with constructed networks attain the same
convergence rate as with the sparse networks with parameters in $[-1,1]$.
| stat |
Vertex Nomination Via Seeded Graph Matching | Consider two networks on overlapping, non-identical vertex sets. Given
vertices of interest in the first network, we seek to identify the
corresponding vertices, if any exist, in the second network. While in
moderately sized networks graph matching methods can be applied directly to
recover the missing correspondences, herein we present a principled methodology
appropriate for situations in which the networks are too large for brute-force
graph matching. Our methodology identifies vertices in a local neighborhood of
the vertices of interest in the first network that have verifiable
corresponding vertices in the second network. Leveraging these known
correspondences, referred to as seeds, we match the induced subgraphs in each
network generated by the neighborhoods of these verified seeds, and rank the
vertices of the second network in terms of the most likely matches to the
original vertices of interest. We demonstrate the applicability of our
methodology through simulations and real data examples.
| stat |
Accelerating Kernel Classifiers Through Borders Mapping | Support vector machines (SVM) and other kernel techniques represent a family
of powerful statistical classification methods with high accuracy and broad
applicability. Because they use all or a significant portion of the training
data, however, they can be slow, especially for large problems. Piecewise
linear classifiers are similarly versatile, yet have the additional advantages
of simplicity, ease of interpretation and, if the number of component linear
classifiers is not too large, speed. Here we show how a simple, piecewise
linear classifier can be trained from a kernel-based classifier in order to
improve the classification speed. The method works by finding the root of the
difference in conditional probabilities between pairs of opposite classes to
build up a representation of the decision boundary. When tested on 17 different
datasets, it succeeded in improving the classification speed of a SVM for 12 of
them by up to two orders-of-magnitude. Of these, two were less accurate than a
simple, linear classifier. The method is best suited to problems with continuum
features data and smooth probability functions. Because the component linear
classifiers are built up individually from an existing classifier, rather than
through a simultaneous optimization procedure, the classifier is also fast to
train.
| stat |
A latent variable approach to account for correlated inputs in global
sensitivity analysis with cases from pharmacological systems modelling | In pharmaceutical research and development decision-making related to drug
candidate selection, efficacy and safety is commonly supported through
modelling and simulation (M\&S). Among others, physiologically-based
pharmacokinetic models are used to describe drug absorption, distribution and
metabolism in human. Global sensitivity analysis (GSA) is gaining interest in
the pharmacological M\&S community as an important element for quality
assessment of model-based inference. Physiological models often present
inter-correlated parameters. The inclusion of correlated factors in GSA and the
sensitivity indices interpretation has proven an issue for these models. Here
we devise and evaluate a latent variable approach for dealing with correlated
factors in GSA. This approach describes the correlation between two model
inputs through the causal relationship of three independent factors: the latent
variable and the unique variances of the two correlated parameters. Then, GSA
is performed with the classical variance-based method. We applied the latent
variable approach to a set of algebraic models and a case from
physiologically-based pharmacokinetics. Then, we compared our approach to
Sobol's GSA assuming no correlations, Sobol's GSA with groups and the
Kucherenko approach. The relative ease of implementation and interpretation
makes this a simple approach for carrying out GSA for models with correlated
input factors.
| stat |
Bayesian analysis of population health data | The analysis of population-wide datasets can provide insight on the health
status of large populations so that public health officials can make
data-driven decisions. The analysis of such datasets often requires highly
parameterized models with different types of fixed and randoms effects to
account for risk factors, spatial and temporal variations, multilevel effects
and other sources on uncertainty. To illustrate the potential of Bayesian
hierarchical models, a dataset of about 500 000 inhabitants released by the
Polish National Health Fund containing information about ischemic stroke
incidence for a 2-year period is analyzed using different types of models.
Spatial logistic regression and survival models are considered for analyzing
the individual probabilities of stroke and the times to the occurrence of an
ischemic stroke event. Demographic and socioeconomic variables as well as drug
prescription information are available at an individual level. Spatial
variation is considered by means of region-level random effects.
| stat |
On the uniqueness and stability of dictionaries for sparse
representation of noisy signals | Learning optimal dictionaries for sparse coding has exposed characteristic
sparse features of many natural signals. However, universal guarantees of the
stability of such features in the presence of noise are lacking. Here, we
provide very general conditions guaranteeing when dictionaries yielding the
sparsest encodings are unique and stable with respect to measurement or
modeling error. We demonstrate that some or all original dictionary elements
are recoverable from noisy data even if the dictionary fails to satisfy the
spark condition, its size is overestimated, or only a polynomial number of
distinct sparse supports appear in the data. Importantly, we derive these
guarantees without requiring any constraints on the recovered dictionary beyond
a natural upper bound on its size. Our results also yield an effective
procedure sufficient to affirm if a proposed solution to the dictionary
learning problem is unique within bounds commensurate with the noise. We
suggest applications to data analysis, engineering, and neuroscience and close
with some remaining challenges left open by our work.
| stat |
Normalization effects on shallow neural networks and related asymptotic
expansions | We consider shallow (single hidden layer) neural networks and characterize
their performance when trained with stochastic gradient descent as the number
of hidden units $N$ and gradient descent steps grow to infinity. In particular,
we investigate the effect of different scaling schemes, which lead to different
normalizations of the neural network, on the network's statistical output,
closing the gap between the $1/\sqrt{N}$ and the mean-field $1/N$
normalization. We develop an asymptotic expansion for the neural network's
statistical output pointwise with respect to the scaling parameter as the
number of hidden units grows to infinity. Based on this expansion, we
demonstrate mathematically that to leading order in $N$, there is no
bias-variance trade off, in that both bias and variance (both explicitly
characterized) decrease as the number of hidden units increases and time grows.
In addition, we show that to leading order in $N$, the variance of the neural
network's statistical output decays as the implied normalization by the scaling
parameter approaches the mean field normalization. Numerical studies on the
MNIST and CIFAR10 datasets show that test and train accuracy monotonically
improve as the neural network's normalization gets closer to the mean field
normalization.
| stat |
Classification of Computer Models with Labelled Outputs | Classification is a vital tool that is important for modelling many complex
numerical models. A model or system may be such that, for certain areas of
input space, the output either does not exist, or is not in a quantifiable
form. Here, we present a new method for classification where the model outputs
are given distinct classifying labels, which we model using a latent Gaussian
process (GP). The latent variable is estimated using MCMC sampling, a unique
likelihood and distinct prior specifications. Our classifier is then verified
by calculating a misclassification rate across the input space.
Comparisons are made with other existing classification methods including
logistic regression, which models the probability of being classified into one
of two regions. To make classification predictions we draw from an independent
Bernoulli distribution, meaning that distance correlation is lost from the
independent draws and so can result in many misclassifications. By modelling
the labels using a latent GP, this problem does not occur in our method. We
apply our novel method to a range of examples including a motivating example
which models the hormones associated with the reproductive system in mammals,
where the two labelled outputs are high and low rates of reproduction.
| stat |
CodedReduce: A Fast and Robust Framework for Gradient Aggregation in
Distributed Learning | We focus on the commonly used synchronous Gradient Descent paradigm for
large-scale distributed learning, for which there has been a growing interest
to develop efficient and robust gradient aggregation strategies that overcome
two key system bottlenecks: communication bandwidth and stragglers' delays. In
particular, Ring-AllReduce (RAR) design has been proposed to avoid bandwidth
bottleneck at any particular node by allowing each worker to only communicate
with its neighbors that are arranged in a logical ring. On the other hand,
Gradient Coding (GC) has been recently proposed to mitigate stragglers in a
master-worker topology by allowing carefully designed redundant allocation of
the data set to the workers. We propose a joint communication topology design
and data set allocation strategy, named CodedReduce (CR), that combines the
best of both RAR and GC. That is, it parallelizes the communications over a
tree topology leading to efficient bandwidth utilization, and carefully designs
a redundant data set allocation and coding strategy at the nodes to make the
proposed gradient aggregation scheme robust to stragglers. In particular, we
quantify the communication parallelization gain and resiliency of the proposed
CR scheme, and prove its optimality when the communication topology is a
regular tree. Furthermore, we empirically evaluate the performance of our
proposed CR design over Amazon EC2 and demonstrate that it achieves speedups of
up to 27.2x and 7.0x, respectively over the benchmarks GC and RAR.
| stat |
High-dimensional Bayesian Optimization of Personalized Cardiac Model
Parameters via an Embedded Generative Model | The estimation of patient-specific tissue properties in the form of model
parameters is important for personalized physiological models. However, these
tissue properties are spatially varying across the underlying anatomical model,
presenting a significance challenge of high-dimensional (HD) optimization at
the presence of limited measurement data. A common solution to reduce the
dimension of the parameter space is to explicitly partition the anatomical
mesh, either into a fixed small number of segments or a multi-scale hierarchy.
This anatomy-based reduction of parameter space presents a fundamental
bottleneck to parameter estimation, resulting in solutions that are either too
low in resolution to reflect tissue heterogeneity, or too high in dimension to
be reliably estimated within feasible computation. In this paper, we present a
novel concept that embeds a generative variational auto-encoder (VAE) into the
objective function of Bayesian optimization, providing an implicit
low-dimensional (LD) search space that represents the generative code of the HD
spatially-varying tissue properties. In addition, the VAE-encoded knowledge
about the generative code is further used to guide the exploration of the
search space. The presented method is applied to estimating tissue excitability
in a cardiac electrophysiological model. Synthetic and real-data experiments
demonstrate its ability to improve the accuracy of parameter estimation with
more than 10x gain in efficiency.
| stat |
Using clinical trial registries to inform Copas selection model for
publication bias in meta-analysis | Prospective registration of study protocols in clinical trial registries is a
useful way to minimize the risk of publication bias in meta-analysis, and
several clinical trial registries are available nowadays. However, they are
mainly used as a tool for searching studies and information submitted to the
registries has not been utilized as efficiently as it could. In addressing
publication bias in meta-analyses, sensitivity analysis with the Copas
selection model is a more objective alternative to widely-used graphical
methods such as the funnel-plot and the trim-and-fill method. Despite its
ability to quantify the potential impact of publication bias, a drawback of the
model is that some parameters not to be specified. This may result in some
difficulty in interpreting the results of the sensitivity analysis. In this
paper, we propose an alternative inference procedure for the Copas selection
model by utilizing information from clinical trial registries. Our method
provides a simple and accurate way to estimate all unknown parameters in the
Copas selection model. A simulation study revealed that our proposed method
resulted in smaller biases and more accurate confidence intervals than existing
methods. Furthermore, two published meta-analyses had been re-analysed to
demonstrate how to implement the proposed method in practice.
| stat |
Gaussian Process Model for Estimating Piecewise Continuous Regression
Functions | This paper presents a Gaussian process (GP) model for estimating piecewise
continuous regression functions. In scientific and engineering applications of
regression analysis, the underlying regression functions are piecewise
continuous in that data follow different continuous regression models for
different regions of the data with possible discontinuities between the
regions. However, many conventional GP regression approaches are not designed
for piecewise regression analysis. We propose a new GP modeling approach for
estimating an unknown piecewise continuous regression function. The new GP
model seeks for a local GP estimate of an unknown regression function at each
test location, using local data neighboring to the test location. To
accommodate the possibilities of the local data from different regions, the
local data is partitioned into two sides by a local linear boundary, and only
the local data belonging to the same side as the test location is used for the
regression estimate. This local split works very well when the input regions
are bounded by smooth boundaries, so the local linear approximation of the
smooth boundaries works well. We estimate the local linear boundary jointly
with the other hyperparameters of the GP model, using the maximum likelihood
approach. Its computation time is as low as the local GP's time. The superior
numerical performance of the proposed approach over the conventional GP
modeling approaches is shown using various simulated piecewise regression
functions.
| stat |
Divergence-based robust inference under proportional hazards model for
one-shot device life-test | In this paper, we develop robust estimators and tests for one-shot device
testing under proportional hazards assumption based on divergence measures.
Through a detailed Monte Carlo simulation study and a numerical example, the
developed inferential procedures are shown to be more robust than the classical
procedures, based on maximum likelihood estimators.
| stat |
Ordering Dimensions with Nested Dropout Normalizing Flows | The latent space of normalizing flows must be of the same dimensionality as
their output space. This constraint presents a problem if we want to learn
low-dimensional, semantically meaningful representations. Recent work has
provided compact representations by fitting flows constrained to manifolds, but
hasn't defined a density off that manifold. In this work we consider flows with
full support in data space, but with ordered latent variables. Like in PCA, the
leading latent dimensions define a sequence of manifolds that lie close to the
data. We note a trade-off between the flow likelihood and the quality of the
ordering, depending on the parameterization of the flow.
| stat |
High Order Adjusted Block-wise Empirical Likelihood For Weakly Dependent
Data | The upper limit on the coverage probability of the empirical likelihood ratio
confidence region severely hampers its application in statistical inferences.
The root cause of this upper limit is the convex hull of the estimating
functions that is used in the construction of the profile empirical likelihood.
For i.i.d data, various methods have been proposed to solve this issue by
modifying the convex hull, but it is not clear how well these methods perform
when the data is no longer independent. In this paper, we consider weakly
dependent multivariate data, and we combine the block-wise empirical likelihood
with the adjusted empirical likelihood to tackle data dependency and the convex
hull constraint simultaneously. We show that our method not only preserves the
much celebrated asymptotic $\chi^2-$distribution, but also improves the
coverage probability by removing the upper limit. Further, we show that our
method is also Bartlett correctable, thus is able to achieve high order
asymptotic coverage accuracy.
| stat |
Discriminator Rejection Sampling | We propose a rejection sampling scheme using the discriminator of a GAN to
approximately correct errors in the GAN generator distribution. We show that
under quite strict assumptions, this will allow us to recover the data
distribution exactly. We then examine where those strict assumptions break down
and design a practical algorithm - called Discriminator Rejection Sampling
(DRS) - that can be used on real data-sets. Finally, we demonstrate the
efficacy of DRS on a mixture of Gaussians and on the SAGAN model,
state-of-the-art in the image generation task at the time of developing this
work. On ImageNet, we train an improved baseline that increases the Inception
Score from 52.52 to 62.36 and reduces the Frechet Inception Distance from 18.65
to 14.79. We then use DRS to further improve on this baseline, improving the
Inception Score to 76.08 and the FID to 13.75.
| stat |
Some Doubly and Multiply Robust Estimators of Controlled Direct Effects | This letter introduces several doubly, triply, and quadruply robust
estimators of the controlled direct effect. Among them, the triply and
quadruply robust estimators are locally semiparametric efficient, and well
suited to the use of data-adaptive methods for estimating their nuisance
functions.
| stat |
Nonstationarity Analysis of Materials Microstructures via Fisher Score
Vectors | Microstructures are critical to the physical properties of materials.
Stochastic microstructures are commonly observed in many kinds of materials and
traditional descriptor-based image analysis of them can be challenging. In this
paper, we introduce a powerful and versatile score-based framework for
analyzing nonstationarity in stochastic materials microstructures. The
framework involves training a parametric supervised learning model to predict a
pixel value using neighboring pixels in images of microstructures~(as known as
micrographs), and this predictive model provides an implicit characterization
of the stochastic nature of the microstructure. The basis for our approach is
the Fisher score vector, defined as the gradient of the log-likelihood with
respect to the parameters of the predictive model, at each micrograph pixel. A
fundamental property of the score vector is that it is zero-mean if the
predictive relationship in the vicinity of that pixel remains unchanged, which
we equate with the local stochastic nature of the microstructure remaining
unchanged. Conversely, if the local stochastic nature changes, then the mean of
the score vector generally differs from zero. Our framework analyzes how the
local mean of the score vector varies across one or more image samples to: (1)
monitor for nonstationarity by indicating whether new samples are statistically
different than reference samples and where they may differ and (2) diagnose
nonstationarity by identifying the distinct types of stochastic microstructures
and labeling accordingly the corresponding regions of the samples. Unlike
feature-based methods, our approach is almost completely general and requires
no prior knowledge of the nature of the nonstationarities. Using a number of
real and simulated micrographs, including polymer composites and multiphase
alloys, we demonstrate the power and versatility of the approach.
| stat |
A unified approach for inference on algorithm-agnostic variable
importance | In many applications, it is of interest to assess the relative contribution
of features (or subsets of features) toward the goal of predicting a response
-- in other words, to gauge the variable importance of features. Most recent
work on variable importance assessment has focused on describing the importance
of features within the confines of a given prediction algorithm. However, such
assessment does not necessarily characterize the prediction potential of
features, and may provide a misleading reflection of the intrinsic value of
these features. To address this limitation, we propose a general framework for
nonparametric inference on interpretable algorithm-agnostic variable
importance. We define variable importance as a population-level contrast
between the oracle predictiveness of all available features versus all features
except those under consideration. We propose a nonparametric efficient
estimation procedure that allows the construction of valid confidence
intervals, even when machine learning techniques are used. We also outline a
valid strategy for testing the null importance hypothesis. Through simulations,
we show that our proposal has good operating characteristics, and we illustrate
its use with data from a study of an antibody against HIV-1 infection.
| stat |
Causal discovery in heavy-tailed models | Causal questions are omnipresent in many scientific problems. While much
progress has been made in the analysis of causal relationships between random
variables, these methods are not well suited if the causal mechanisms only
manifest themselves in extremes. This work aims to connect the two fields of
causal inference and extreme value theory. We define the causal tail
coefficient that captures asymmetries in the extremal dependence of two random
variables. In the population case, the causal tail coefficient is shown to
reveal the causal structure if the distribution follows a linear structural
causal model. This holds even in the presence of latent common causes that have
the same tail index as the observed variables. Based on a consistent estimator
of the causal tail coefficient, we propose a computationally highly efficient
algorithm that estimates the causal structure. We prove that our method
consistently recovers the causal order and we compare it to other
well-established and non-extremal approaches in causal discovery on synthetic
and real data. The code is available as an open-access R package.
| stat |
On the role of surrogates in the efficient estimation of treatment
effects with limited outcome data | We study the problem of estimating treatment effects when the outcome of
primary interest (e.g., long-term health status) is only seldom observed but
abundant surrogate observations (e.g., short-term health outcomes) are
available. To investigate the role of surrogates in this setting, we derive the
semiparametric efficiency lower bounds of average treatment effect (ATE) both
with and without presence of surrogates, as well as several intermediary
settings. These bounds characterize the best-possible precision of ATE
estimation in each case, and their difference quantifies the efficiency gains
from optimally leveraging the surrogates in terms of key problem
characteristics when only limited outcome data are available. We show these
results apply in two important regimes: when the number of surrogate
observations is comparable to primary-outcome observations and when the former
dominates the latter. Importantly, we take a missing-data approach that
circumvents strong surrogate conditions which are commonly assumed in previous
literature but almost always fail in practice. To show how to leverage the
efficiency gains of surrogate observations, we propose ATE estimators and
inferential methods based on flexible machine learning methods to estimate
nuisance parameters that appear in the influence functions. We show our
estimators enjoy efficiency and robustness guarantees under weak conditions.
| stat |
Mean-field theory of two-layers neural networks: dimension-free bounds
and kernel limit | We consider learning two layer neural networks using stochastic gradient
descent. The mean-field description of this learning dynamics approximates the
evolution of the network weights by an evolution in the space of probability
distributions in $R^D$ (where $D$ is the number of parameters associated to
each neuron). This evolution can be defined through a partial differential
equation or, equivalently, as the gradient flow in the Wasserstein space of
probability distributions. Earlier work shows that (under some regularity
assumptions), the mean field description is accurate as soon as the number of
hidden units is much larger than the dimension $D$. In this paper we establish
stronger and more general approximation guarantees. First of all, we show that
the number of hidden units only needs to be larger than a quantity dependent on
the regularity properties of the data, and independent of the dimensions. Next,
we generalize this analysis to the case of unbounded activation functions,
which was not covered by earlier bounds. We extend our results to noisy
stochastic gradient descent.
Finally, we show that kernel ridge regression can be recovered as a special
limit of the mean field analysis.
| stat |
Gaussian Process Assisted Active Learning of Physical Laws | In many areas of science and engineering, discovering the governing
differential equations from the noisy experimental data is an essential
challenge. It is also a critical step in understanding the physical phenomena
and prediction of the future behaviors of the systems. However, in many cases,
it is expensive or time-consuming to collect experimental data. This article
provides an active learning approach to estimate the unknown differential
equations accurately with reduced experimental data size. We propose an
adaptive design criterion combining the D-optimality and the maximin
space-filling criterion. In contrast to active learning for other regression
models, the D-optimality here requires the unknown solution of the differential
equations and derivatives of the solution. We estimate the Gaussian process
(GP) regression models from the available experimental data and use them as the
surrogates of these unknown solution functions. The derivatives of the
estimated GP models are derived and used to substitute the derivatives of the
solution. Variable selection-based regression methods are used to learn the
differential equations from the experimental data. Through multiple case
studies, we demonstrate the proposed approach outperforms the D-optimality and
the maximin space-filling design alone in terms of model accuracy and data
economy.
| stat |
Stacked Capsule Autoencoders | Objects are composed of a set of geometrically organized parts. We introduce
an unsupervised capsule autoencoder (SCAE), which explicitly uses geometric
relationships between parts to reason about objects. Since these relationships
do not depend on the viewpoint, our model is robust to viewpoint changes. SCAE
consists of two stages. In the first stage, the model predicts presences and
poses of part templates directly from the image and tries to reconstruct the
image by appropriately arranging the templates. In the second stage, SCAE
predicts parameters of a few object capsules, which are then used to
reconstruct part poses. Inference in this model is amortized and performed by
off-the-shelf neural encoders, unlike in previous capsule networks. We find
that object capsule presences are highly informative of the object class, which
leads to state-of-the-art results for unsupervised classification on SVHN (55%)
and MNIST (98.7%). The code is available at
https://github.com/google-research/google-research/tree/master/stacked_capsule_autoencoders
| stat |
Estimation in the Cox Survival Regression Model with Covariate
Measurement Error and a Changepoint | The Cox regression model is a popular model for analyzing the relationship
between a covariate and a survival endpoint. The standard Cox model assumes a
constant covariate effect across the entire covariate domain. However, in many
epidemiological and other applications, the covariate of main interest is
subject to a threshold effect: a change in the slope at a certain point within
the covariate domain. Often, the covariate of interest is subject to some
degree of measurement error. In this paper, we study measurement error
correction in the case where the threshold is known. Several bias correction
methods are examined: two versions of regression calibration (RC1 and RC2, the
latter of which is new), two methods based on the induced relative risk under a
rare event assumption (RR1 and RR2, the latter of which is new), a maximum
pseudo-partial likelihood estimator (MPPLE), and simulation-extrapolation
(SIMEX). We develop the theory, present simulations comparing the methods, and
illustrate their use on data concerning the relationship between chronic air
pollution exposure to particulate matter PM10 and fatal myocardial infarction
(Nurses Health Study (NHS)), and on data concerning the effect of a subject's
long-term underlying systolic blood pressure level on the risk of
cardiovascular disease death (Framingham Heart Study (FHS)). The simulations
indicate that the best methods are RR2 and MPPLE.
| stat |
Numerical computation of triangular complex spherical designs with small
mesh ratio | This paper provides triangular spherical designs for the complex unit sphere
$\Omega^d$ by exploiting the natural correspondence between the complex unit
sphere in $d$ dimensions and the real unit sphere in $2d-1$. The existence of
triangular and square complex spherical $t$-designs with the optimal order
number of points is established. A variational characterization of triangular
complex designs is provided, with particular emphasis on numerical computation
of efficient triangular complex designs with good geometric properties as
measured by their mesh ratio. We give numerical examples of triangular
spherical $t$-designs on complex unit spheres of dimension $d=2$ to $6$.
| stat |
Inverse probability of censoring weighting for visual predictive checks
of time-to-event models with time-varying covariates | When constructing models to summarize clinical data to be used for
simulations, it is good practice to evaluate the models for their capacity to
reproduce the data. This can be done by means of Visual Predictive Checks
(VPC), which consist of (1) several reproductions of the original study by
simulation from the model under evaluation, (2) calculating estimates of
interest for each simulated study and (3) comparing the distribution of those
estimates with the estimate from the original study. This procedure is a
generic method that is straightforward to apply, in general. Here we consider
the application of the method to time to event data and consider the special
case when a time-varying covariate is not known or cannot be approximated after
event time. In this case, simulations cannot be conducted beyond the end of the
follow-up time (event or censoring time) in the original study. Thus, the
simulations must be censored at the end of the follow-up time. Since this
censoring is not random, the standard KM estimates from the simulated studies
and the resulting VPC will be biased. We propose to use inverse probability of
censoring weighting (IPoC) method to correct the KM estimator for the simulated
studies and obtain unbiased VPCs. For analyzing the Cantos study, the IPoC
weighting as described here proved valuable and enabled the generation of VPCs
to qualify PKPD models for simulations. Here, we use a generated data set,
which allows illustration of the different situations and evaluation against
the known truth.
| stat |
A Comparison of Aggregation Methods for Probabilistic Forecasts of
COVID-19 Mortality in the United States | The COVID-19 pandemic has placed forecasting models at the forefront of
health policy making. Predictions of mortality and hospitalization help
governments meet planning and resource allocation challenges. In this paper, we
consider the weekly forecasting of the cumulative mortality due to COVID-19 at
the national and state level in the U.S. Optimal decision-making requires a
forecast of a probability distribution, rather than just a single point
forecast. Interval forecasts are also important, as they can support decision
making and provide situational awareness. We consider the case where
probabilistic forecasts have been provided by multiple forecasting teams, and
we aggregate the forecasts to extract the wisdom of the crowd. With only
limited information available regarding the historical accuracy of the
forecasting teams, we consider aggregation (i.e. combining) methods that do not
rely on a record of past accuracy. In this empirical paper, we evaluate the
accuracy of aggregation methods that have been previously proposed for interval
forecasts and predictions of probability distributions. These include the use
of the simple average, the median, and trimming methods, which enable robust
estimation and allow the aggregate forecast to reduce the impact of a tendency
for the forecasting teams to be under- or overconfident. We use data that has
been made publicly available from the COVID-19 Forecast Hub. While the simple
average performed well for the high mortality series, we obtained greater
accuracy using the median and certain trimming methods for the low and medium
mortality series. It will be interesting to see if this remains the case as the
pandemic evolves.
| stat |
High-dimensional single-index Bayesian modeling of brain atrophy | We propose a model of brain atrophy as a function of high-dimensional genetic
information and low dimensional covariates such as gender, age, APOE gene, and
disease status. A nonparametric single-index Bayesian model of high dimension
is proposed to model the relationship with B-spline series prior on the unknown
functions and Dirichlet process scale mixture of centered normal prior on the
distributions of the random effects. The posterior rate of contraction without
the random effect is established for a fixed number of regions and time points
with increasing sample size. We implement an efficient computation algorithm
through a Hamiltonian Monte Carlo (HMC) algorithm. The performance of the
proposed Bayesian method is compared with the corresponding least square
estimator in the linear model with horseshoe prior, LASSO and SCAD penalization
on the high-dimensional covariates. The proposed Bayesian method is applied to
a dataset on volumes of brain regions recorded over multiple visits of 748
individuals using 620,901 SNPs and 6 other covariates for each individual, to
identify factors associated with brain atrophy.
| stat |
Modelling High-Dimensional Categorical Data Using Nonconvex Fusion
Penalties | We propose a method for estimation in high-dimensional linear models with
nominal categorical data. Our estimator, called SCOPE, fuses levels together by
making their corresponding coefficients exactly equal. This is achieved using
the minimax concave penalty on differences between the order statistics of the
coefficients for a categorical variable, thereby clustering the coefficients.
We provide an algorithm for exact and efficient computation of the global
minimum of the resulting nonconvex objective in the case with a single variable
with potentially many levels, and use this within a block coordinate descent
procedure in the multivariate case. We show that an oracle least squares
solution that exploits the unknown level fusions is a limit point of the
coordinate descent with high probability, provided the true levels have a
certain minimum separation; these conditions are known to be minimal in the
univariate case. We demonstrate the favourable performance of SCOPE across a
range of real and simulated datasets. An R package CatReg implementing SCOPE
for linear models and also a version for logistic regression is available on
CRAN.
| stat |
Optimizing regularized Cholesky score for order-based learning of
Bayesian networks | Bayesian networks are a class of popular graphical models that encode causal
and conditional independence relations among variables by directed acyclic
graphs (DAGs). We propose a novel structure learning method, annealing on
regularized Cholesky score (ARCS), to search over topological sorts, or
permutations of nodes, for a high-scoring Bayesian network. Our scoring
function is derived from regularizing Gaussian DAG likelihood, and its
optimization gives an alternative formulation of the sparse Cholesky
factorization problem from a statistical viewpoint, which is of independent
interest. We combine global simulated annealing over permutations with a fast
proximal gradient algorithm, operating on triangular matrices of edge
coefficients, to compute the score of any permutation. Combined, the two
approaches allow us to quickly and effectively search over the space of DAGs
without the need to verify the acyclicity constraint or to enumerate possible
parent sets given a candidate topological sort. The annealing aspect of the
optimization is able to consistently improve the accuracy of DAGs learned by
local search algorithms. In addition, we develop several techniques to
facilitate the structure learning, including pre-annealing data-driven tuning
parameter selection and post-annealing constraint-based structure refinement.
Through extensive numerical comparisons, we show that ARCS achieves substantial
improvements over existing methods, demonstrating its great potential to learn
Bayesian networks from both observational and experimental data.
| stat |
Social Distancing and COVID-19: Randomization Inference for a Structured
Dose-Response Relationship | Social distancing is widely acknowledged as an effective public health policy
combating the novel coronavirus. But extreme social distancing has costs and it
is not clear how much social distancing is needed to achieve public health
effects. In this article, we develop a design-based framework to make inference
about the dose-response relationship between social distancing and COVID-19
related death toll and case numbers. We first discuss how to embed
observational data with a time-independent, continuous treatment dose into an
approximate randomized experiment, and develop a randomization-based procedure
that tests if a structured dose-response relationship fits the data. We then
generalize the design and testing procedure to accommodate a time-dependent,
treatment dose trajectory, and generalize a dose-response relationship to a
longitudinal setting. Finally, we apply the proposed design and testing
procedures to investigate the effect of social distancing during the phased
reopening in the United States on public health outcomes using data compiled
from sources including Unacast, the United States Census Bureau, and the County
Health Rankings and Roadmaps Program. We test a primary analysis hypothesis
that states the social distancing from April 27th to June 28th had no effect on
the COVID-19-related death toll from June 29th to August 2nd (p-value < 0.001)
and conducted extensive secondary analyses that investigate the dose-response
relationship between social distancing and COVID-19 case numbers.
| stat |
BVAR-Connect: A Variational Bayes Approach to Multi-Subject Vector
Autoregressive Models for Inference on Brain Connectivity Networks | In this paper we propose BVAR-connect, a variational inference approach to a
Bayesian multi-subject vector autoregressive (VAR) model for inference on
effective brain connectivity based on resting-state functional MRI data. The
modeling framework uses a Bayesian variable selection approach that flexibly
integrates multi-modal data, in particular structural diffusion tensor imaging
(DTI) data, into the prior construction. The variational inference approach we
develop allows scalability of the methods and results in the ability to
estimate subject- and group-level brain connectivity networks over whole-brain
parcellations of the data. We provide a brief description of a user-friendly
MATLAB GUI released for public use. We assess performance on simulated data,
where we show that the proposed inference method can achieve comparable
accuracy to the sampling-based Markov Chain Monte Carlo approach but at a much
lower computational cost. We also address the case of subject groups with
imbalanced sample sizes. Finally, we illustrate the methods on resting-state
functional MRI and structural DTI data on children with a history of traumatic
injury.
| stat |
G-Formula for Observational Studies with Partial Interference, with
Application to Bed Net Use on Malaria | Assessing population-level effects of vaccines and other infectious disease
prevention measures is important to the field of public health. In infectious
disease studies, one person's treatment may affect another individual's
outcome, i.e., there may be interference between units. For example, use of bed
nets to prevent malaria by one individual may have an indirect or spillover
effect to other individuals living in close proximity. In some settings,
individuals may form groups or clusters where interference only occurs within
groups, i.e., there is partial interference. Inverse probability weighted
estimators have previously been developed for observational studies with
partial interference. Unfortunately, these estimators are not well suited for
studies with large clusters. Therefore, in this paper, the parametric g-formula
is extended to allow for partial interference. G-formula estimators are
proposed of overall effects, spillover effects when treated, and spillover
effects when untreated. The proposed estimators can accommodate large clusters
and do not suffer from the g-null paradox that may occur in the absence of
interference. The large sample properties of the proposed estimators are
derived, and simulation studies are presented demonstrating the finite-sample
performance of the proposed estimators. The Demographic and Health Survey from
the Democratic Republic of the Congo is then analyzed using the proposed
g-formula estimators to assess the overall and spillover effects of bed net use
on malaria.
| stat |
Building Calibrated Deep Models via Uncertainty Matching with Auxiliary
Interval Predictors | With rapid adoption of deep learning in critical applications, the question
of when and how much to trust these models often arises, which drives the need
to quantify the inherent uncertainties. While identifying all sources that
account for the stochasticity of models is challenging, it is common to augment
predictions with confidence intervals to convey the expected variations in a
model's behavior. We require prediction intervals to be well-calibrated,
reflect the true uncertainties, and to be sharp. However, existing techniques
for obtaining prediction intervals are known to produce unsatisfactory results
in at least one of these criteria. To address this challenge, we develop a
novel approach for building calibrated estimators. More specifically, we use
separate models for prediction and interval estimation, and pose a bi-level
optimization problem that allows the former to leverage estimates from the
latter through an \textit{uncertainty matching} strategy. Using experiments in
regression, time-series forecasting, and object localization, we show that our
approach achieves significant improvements over existing uncertainty
quantification methods, both in terms of model fidelity and calibration error.
| stat |
The Benefits of Over-parameterization at Initialization in Deep ReLU
Networks | It has been noted in existing literature that over-parameterization in ReLU
networks generally improves performance. While there could be several factors
involved behind this, we prove some desirable theoretical properties at
initialization which may be enjoyed by ReLU networks. Specifically, it is known
that He initialization in deep ReLU networks asymptotically preserves variance
of activations in the forward pass and variance of gradients in the backward
pass for infinitely wide networks, thus preserving the flow of information in
both directions. Our paper goes beyond these results and shows novel properties
that hold under He initialization: i) the norm of hidden activation of each
layer is equal to the norm of the input, and, ii) the norm of weight gradient
of each layer is equal to the product of norm of the input vector and the error
at output layer. These results are derived using the PAC analysis framework,
and hold true for finitely sized datasets such that the width of the ReLU
network only needs to be larger than a certain finite lower bound. As we show,
this lower bound depends on the depth of the network and the number of samples,
and by the virtue of being a lower bound, over-parameterized ReLU networks are
endowed with these desirable properties. For the aforementioned hidden
activation norm property under He initialization, we further extend our theory
and show that this property holds for a finite width network even when the
number of data samples is infinite. Thus we overcome several limitations of
existing papers, and show new properties of deep ReLU networks at
initialization.
| stat |
Learning higher-order sequential structure with cloned HMMs | Variable order sequence modeling is an important problem in artificial and
natural intelligence. While overcomplete Hidden Markov Models (HMMs), in
theory, have the capacity to represent long-term temporal structure, they often
fail to learn and converge to local minima. We show that by constraining HMMs
with a simple sparsity structure inspired by biology, we can make it learn
variable order sequences efficiently. We call this model cloned HMM (CHMM)
because the sparsity structure enforces that many hidden states map
deterministically to the same emission state. CHMMs with over 1 billion
parameters can be efficiently trained on GPUs without being severely affected
by the credit diffusion problem of standard HMMs. Unlike n-grams and sequence
memoizers, CHMMs can model temporal dependencies at arbitrarily long distances
and recognize contexts with 'holes' in them. Compared to Recurrent Neural
Networks and their Long Short-Term Memory extensions (LSTMs), CHMMs are
generative models that can natively deal with uncertainty. Moreover, CHMMs
return a higher-order graph that represents the temporal structure of the data
which can be useful for community detection, and for building hierarchical
models. Our experiments show that CHMMs can beat n-grams, sequence memoizers,
and LSTMs on character-level language modeling tasks. CHMMs can be a viable
alternative to these methods in some tasks that require variable order sequence
modeling and the handling of uncertainty.
| stat |
Improving Spectral Clustering using the Asymptotic Value of the
Normalised Cut | Spectral clustering is a popular and versatile clustering method based on a
relaxation of the normalised graph cut objective. Despite its popularity,
however, there is no single agreed upon method for tuning the important scaling
parameter, nor for determining automatically the number of clusters to extract.
Popular heuristics exist, but corresponding theoretical results are scarce. In
this paper we investigate the asymptotic value of the normalised cut for an
increasing sample assumed to arise from an underlying probability distribution,
and based on this result provide recommendations for improving spectral
clustering methodology. A corresponding algorithm is proposed with strong
empirical performance.
| stat |
casebase: An Alternative Framework For Survival Analysis and Comparison
of Event Rates | In epidemiological studies of time-to-event data, a quantity of interest to
the clinician and the patient is the risk of an event given a covariate
profile. However, methods relying on time matching or risk-set sampling
(including Cox regression) eliminate the baseline hazard from the likelihood
expression or the estimating function. The baseline hazard then needs to be
estimated separately using a non-parametric approach. This leads to step-wise
estimates of the cumulative incidence that are difficult to interpret. Using
case-base sampling, Hanley & Miettinen (2009) explained how the parametric
hazard functions can be estimated using logistic regression. Their approach
naturally leads to estimates of the cumulative incidence that are
smooth-in-time. In this paper, we present the casebase R package, a
comprehensive and flexible toolkit for parametric survival analysis. We
describe how the case-base framework can also be used in more complex settings:
competing risks, time-varying exposure, and variable selection. Our package
also includes an extensive array of visualization tools to complement the
analysis of time-to-event data. We illustrate all these features through four
different case studies. *SRB and MT contributed equally to this work.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.