title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Canonical Least Favorable Submodels:A New TMLE Procedure for
Multidimensional Parameters | This paper is a fundamental addition to the world of targeted maximum
likelihood estimation (TMLE) (or likewise, targeted minimum loss estimation)
for simultaneous estimation of multi-dimensional parameters of interest. TMLE,
as part of the targeted learning framework, offers a crucial step in
constructing efficient plug-in estimators for nonparametric or semiparametric
models. The so-called targeting step of targeted learning, involves fluctuating
the initial fit of the model in a way that maximally adjusts the plug-in
estimate per change in the log likelihood. Previously for multidimensional
parameters of interest, iterative TMLE's were constructed using locally least
favorable submodels as defined in van der Laan and Gruber, 2016, which are
indexed by a multidimensional fluctuation parameter. In this paper we define a
canonical least favorable submodel in terms of a single dimensional epsilon for
a $d$-dimensional parameter of interest. One can view the clfm as the iterative
analog to the one-step TMLE as constructed in van der Laan and Gruber, 2016. It
is currently implemented in several software packages we provide in the last
section. Using a single epsilon for the targeting step in TMLE could be useful
for high dimensional parameters, where using a fluctuation parameter of the
same dimension as the parameter of interest could suffer the consequences of
curse of dimensionality. The clfm also enables placing the so-called clever
covariate denominator as an inverse weight in an offset intercept model. It has
been shown that such weighting mitigates the effect of large inverse weights
sometimes caused by near positivity violations.
| stat |
Unsupervised Fault Detection in Varying Operating Conditions | Training data-driven approaches for complex industrial system health
monitoring is challenging. When data on faulty conditions are rare or not
available, the training has to be performed in a unsupervised manner. In
addition, when the observation period, used for training, is kept short, to be
able to monitor the system in its early life, the training data might not be
representative of all the system normal operating conditions. In this paper, we
propose five approaches to perform fault detection in such context. Two
approaches rely on the data from the unit to be monitored only: the baseline is
trained on the early life of the unit. An incremental learning procedure tries
to learn new operating conditions as they arise. Three other approaches take
advantage of data from other similar units within a fleet. In two cases, units
are directly compared to each other with similarity measures, and the data from
similar units are combined in the training set. We propose, in the third case,
a new deep-learning methodology to perform, first, a feature alignment of
different units with an Unsupervised Feature Alignment Network (UFAN). Then,
features of both units are combined in the training set of the fault detection
neural network.
The approaches are tested on a fleet comprising 112 units, observed over one
year of data. All approaches proposed here are an improvement to the baseline,
trained with two months of data only. As units in the fleet are found to be
very dissimilar, the new architecture UFAN, that aligns units in the feature
space, is outperforming others.
| stat |
Identifying predictive biomarkers of CIMAvaxEGF success in advanced Lung
Cancer Patients | Objectives: To identify predictive biomarkers of CIMAvaxEGF success in the
treatment of Non-Small Cell Lung Cancer Patients. Methods: Data from a clinical
trial evaluating the effect on survival time of CIMAvax-EGF versus best
supportive care were analyzed retrospectively following the causal inference
approach. Pre-treatment potential predictive biomarkers included basal serum
EGF concentration, peripheral blood parameters and immunosenescence biomarkers
(The proportion of CD8 + CD28- T cells, CD4+ and CD8+ T cells, CD4 CD8 ratio
and CD19+ B cells. The 33 patients with complete information were included. The
predictive causal information (PCI) was calculated for all possible models. The
model with a minimum number of predictors, but with high prediction accuracy
(PCI>0.7) was selected. Good, rare and poor responder patients were identified
using the predictive probability of treatment success. Results: The mean of PCI
increased from 0.486, when only one predictor is considered, to 0.98 using the
multivariate approach with all predictors. The model considering the proportion
of CD4+ T cell, basal EGF concentration, NLR, Monocytes, and Neutrophils as
predictors were selected (PCI>0.74). Patients predicted as good responders
according to the pre-treatment biomarkers values treated with CIMAvax-EGF had a
significant higher observed survival compared with the control group (p=0.03).
No difference was observed for bad responders. Conclusions: Peripheral blood
parameters and immunosenescence biomarkers together with basal EGF
concentration in serum resulted in good predictors of the CIMAvax-EGF success
in advanced NSCLC. The study illustrates the application of a new methodology,
based on causal inference, to evaluate multivariate pre-treatment predictors
| stat |
Continual Learning with Adaptive Weights (CLAW) | Approaches to continual learning aim to successfully learn a set of related
tasks that arrive in an online manner. Recently, several frameworks have been
developed which enable deep learning to be deployed in this learning scenario.
A key modelling decision is to what extent the architecture should be shared
across tasks. On the one hand, separately modelling each task avoids
catastrophic forgetting but it does not support transfer learning and leads to
large models. On the other hand, rigidly specifying a shared component and a
task-specific part enables task transfer and limits the model size, but it is
vulnerable to catastrophic forgetting and restricts the form of task-transfer
that can occur. Ideally, the network should adaptively identify which parts of
the network to share in a data driven way. Here we introduce such an approach
called Continual Learning with Adaptive Weights (CLAW), which is based on
probabilistic modelling and variational inference. Experiments show that CLAW
achieves state-of-the-art performance on six benchmarks in terms of overall
continual learning performance, as measured by classification accuracy, and in
terms of addressing catastrophic forgetting.
| stat |
Conditional calibration for false discovery rate control under
dependence | We introduce a new class of methods for finite-sample false discovery rate
(FDR) control in multiple testing problems with dependent test statistics where
the dependence is fully or partially known. Our approach separately calibrates
a data-dependent p-value rejection threshold for each hypothesis, relaxing or
tightening the threshold as appropriate to target exact FDR control. In
addition to our general framework we propose a concrete algorithm, the
dependence-adjusted Benjamini-Hochberg (dBH) procedure, which adaptively
thresholds the q-value for each hypothesis. Under positive regression
dependence the dBH procedure uniformly dominates the standard BH procedure, and
in general it uniformly dominates the Benjamini-Yekutieli (BY) procedure (also
known as BH with log correction). Simulations and real data examples illustrate
power gains over competing approaches to FDR control under dependence.
| stat |
BIMC: The Bayesian Inverse Monte Carlo method for goal-oriented
uncertainty quantification. Part I | We consider the problem of estimating rare event probabilities, focusing on
systems whose evolution is governed by differential equations with uncertain
input parameters. If the system dynamics is expensive to compute, standard
sampling algorithms such as the Monte Carlo method may require infeasible
running times to accurately evaluate these probabilities. We propose an
importance sampling scheme (which we call BIMC) that relies on solving an
auxiliary, fictitious Bayesian inverse problem. The solution of the inverse
problem yields a posterior PDF, a local Gaussian approximation to which serves
as the importance sampling density. We apply BIMC to several problems and
demonstrate that it can lead to computational savings of several orders of
magnitude over the Monte Carlo method. We delineate conditions under which BIMC
is optimal, as well as conditions when it can fail to yield an effective IS
density.
| stat |
Sorting Big Data by Revealed Preference with Application to College
Ranking | When ranking big data observations such as colleges in the United States,
diverse consumers reveal heterogeneous preferences. The objective of this paper
is to sort out a linear ordering for these observations and to recommend
strategies to improve their relative positions in the ranking. A properly
sorted solution could help consumers make the right choices, and governments
make wise policy decisions. Previous researchers have applied exogenous
weighting or multivariate regression approaches to sort big data objects,
ignoring their variety and variability. By recognizing the diversity and
heterogeneity among both the observations and the consumers, we instead apply
endogenous weighting to these contradictory revealed preferences. The outcome
is a consistent steady-state solution to the counterbalance equilibrium within
these contradictions. The solution takes into consideration the spillover
effects of multiple-step interactions among the observations. When information
from data is efficiently revealed in preferences, the revealed preferences
greatly reduce the volume of the required data in the sorting process. The
employed approach can be applied in many other areas, such as sports team
ranking, academic journal ranking, voting, and real effective exchange rates.
| stat |
The role of surrogate models in the development of digital twins of
dynamic systems | Digital twin technology has significant promise, relevance and potential of
widespread applicability in various industrial sectors such as aerospace,
infrastructure and automotive. However, the adoption of this technology has
been slower due to the lack of clarity for specific applications. A discrete
damped dynamic system is used in this paper to explore the concept of a digital
twin. As digital twins are also expected to exploit data and computational
methods, there is a compelling case for the use of surrogate models in this
context. Motivated by this synergy, we have explored the possibility of using
surrogate models within the digital twin technology. In particular, the use of
Gaussian process (GP) emulator within the digital twin technology is explored.
GP has the inherent capability of addressing noise and sparse data and hence,
makes a compelling case to be used within the digital twin framework. Cases
involving stiffness variation and mass variation are considered, individually
and jointly along with different levels of noise and sparsity in data. Our
numerical simulation results clearly demonstrate that surrogate models such as
GP emulators have the potential to be an effective tool for the development of
digital twins. Aspects related to data quality and sampling rate are analysed.
Key concepts introduced in this paper are summarised and ideas for urgent
future research needs are proposed.
| stat |
Dynamic Gauss Newton Metropolis Algorithm | GNM: The MCMC Jagger. A rocking awesome sampler. This python package is an
affine invariant Markov chain Monte Carlo (MCMC) sampler based on the dynamic
Gauss-Newton-Metropolis (GNM) algorithm. The GNM algorithm is specialized in
sampling highly non-linear posterior probability distribution functions of the
form $e^{-||f(x)||^2/2}$, and the package is an implementation of this
algorithm. On top of the back-off strategy in the original GNM algorithm, there
is the dynamic hyper-parameter optimization feature added to the algorithm and
included in the package to help increase performance of the back-off and
therefore the sampling. Also, there are the Jacobian tester, error bars creator
and many more features for the ease of use included in the code. The problem is
introduced and a guide to installation is given in the introduction. Then how
to use the python package is explained. The algorithm is given and finally
there are some examples using exponential time series to show the performance
of the algorithm and the back-off strategy.
| stat |
Active Invariant Causal Prediction: Experiment Selection through
Stability | A fundamental difficulty of causal learning is that causal models can
generally not be fully identified based on observational data only.
Interventional data, that is, data originating from different experimental
environments, improves identifiability. However, the improvement depends
critically on the target and nature of the interventions carried out in each
experiment. Since in real applications experiments tend to be costly, there is
a need to perform the right interventions such that as few as possible are
required. In this work we propose a new active learning (i.e. experiment
selection) framework (A-ICP) based on Invariant Causal Prediction (ICP) (Peters
et al., 2016). For general structural causal models, we characterize the effect
of interventions on so-called stable sets, a notion introduced by (Pfister et
al., 2019). We leverage these results to propose several intervention selection
policies for A-ICP which quickly reveal the direct causes of a response
variable in the causal graph while maintaining the error control inherent in
ICP. Empirically, we analyze the performance of the proposed policies in both
population and finite-regime experiments.
| stat |
Localized Structured Prediction | Key to structured prediction is exploiting the problem structure to simplify
the learning process. A major challenge arises when data exhibit a local
structure (e.g., are made by "parts") that can be leveraged to better
approximate the relation between (parts of) the input and (parts of) the
output. Recent literature on signal processing, and in particular computer
vision, has shown that capturing these aspects is indeed essential to achieve
state-of-the-art performance. While such algorithms are typically derived on a
case-by-case basis, in this work we propose the first theoretical framework to
deal with part-based data from a general perspective. We derive a novel
approach to deal with these problems and study its generalization properties
within the setting of statistical learning theory. Our analysis is novel in
that it explicitly quantifies the benefits of leveraging the part-based
structure of the problem with respect to the learning rates of the proposed
estimator.
| stat |
Bayesian Higher Order Hidden Markov Models | We consider the problem of flexible modeling of higher order hidden Markov
models when the number of latent states and the nature of the serial
dependence, including the true order, are unknown. We propose Bayesian
nonparametric methodology based on tensor factorization techniques that can
characterize any transition probability with a specified maximal order,
allowing automated selection of the important lags and capturing higher order
interactions among the lags. Theoretical results provide insights into
identifiability of the emission distributions and asymptotic behavior of the
posterior. We design efficient Markov chain Monte Carlo algorithms for
posterior computation. In simulation experiments, the method vastly
outperformed its first and higher order competitors not just in higher order
settings, but, remarkably, also in first order cases. Practical utility is
illustrated using real world applications.
| stat |
Outcome-Guided Disease Subtyping for High-Dimensional Omics Data | High-throughput microarray and sequencing technology have been used to
identify disease subtypes that could not be observed otherwise by using
clinical variables alone. The classical unsupervised clustering strategy
concerns primarily the identification of subpopulations that have similar
patterns in gene features. However, as the features corresponding to irrelevant
confounders (e.g. gender or age) may dominate the clustering process, the
resulting clusters may or may not capture clinically meaningful disease
subtypes. This gives rise to a fundamental problem: can we find a subtyping
procedure guided by a pre-specified disease outcome? Existing methods, such as
supervised clustering, apply a two-stage approach and depend on an arbitrary
number of selected features associated with outcome. In this paper, we propose
a unified latent generative model to perform outcome-guided disease subtyping
constructed from omics data, which improves the resulting subtypes concerning
the disease of interest. Feature selection is embedded in a regularization
regression. A modified EM algorithm is applied for numerical computation and
parameter estimation. The proposed method performs feature selection, latent
subtype characterization and outcome prediction simultaneously. To account for
possible outliers or violation of mixture Gaussian assumption, we incorporate
robust estimation using adaptive Huber or median-truncated loss function.
Extensive simulations and an application to complex lung diseases with
transcriptomic and clinical data demonstrate the ability of the proposed method
to identify clinically relevant disease subtypes and signature genes suitable
to explore toward precision medicine.
| stat |
Statistical Uncertainty Analysis for Stochastic Simulation | When we use simulation to evaluate the performance of a stochastic system,
the simulation often contains input distributions estimated from real-world
data; therefore, there is both simulation and input uncertainty in the
performance estimates. Ignoring either source of uncertainty underestimates the
overall statistical error. Simulation uncertainty can be reduced by additional
computation (e.g., more replications). Input uncertainty can be reduced by
collecting more real-world data, when feasible. This paper proposes an approach
to quantify overall statistical uncertainty when the simulation is driven by
independent parametric input distributions; specifically, we produce a
confidence interval that accounts for both simulation and input uncertainty by
using a metamodel-assisted bootstrapping approach. The input uncertainty is
measured via bootstrapping, an equation-based stochastic kriging metamodel
propagates the input uncertainty to the output mean, and both simulation and
metamodel uncertainty are derived using properties of the metamodel. A variance
decomposition is proposed to estimate the relative contribution of input to
overall uncertainty; this information indicates whether the overall uncertainty
can be significantly reduced through additional simulation alone. Asymptotic
analysis provides theoretical support for our approach, while an empirical
study demonstrates that it has good finite-sample performance.
| stat |
A Generative Model for Sampling High-Performance and Diverse Weights for
Neural Networks | Recent work on mode connectivity in the loss landscape of deep neural
networks has demonstrated that the locus of (sub-)optimal weight vectors lies
on continuous paths. In this work, we train a neural network that serves as a
hypernetwork, mapping a latent vector into high-performance (low-loss) weight
vectors, generalizing recent findings of mode connectivity to higher
dimensional manifolds. We formulate the training objective as a compromise
between accuracy and diversity, where the diversity takes into account trivial
symmetry transformations of the target network. We demonstrate how to reduce
the number of parameters in the hypernetwork by parameter sharing. Once
learned, the hypernetwork allows for a computationally efficient, ancestral
sampling of neural network weights, which we recruit to form large ensembles.
The improvement in classification accuracy obtained by this ensembling
indicates that the generated manifold extends in dimensions other than
directions implied by trivial symmetries. For computational efficiency, we
distill an ensemble into a single classifier while retaining generalization.
| stat |
The Importance of Being a Band: Finite-Sample Exact Distribution-Free
Prediction Sets for Functional Data | Functional Data Analysis represents a field of growing interest in
statistics. Despite several studies have been proposed leading to fundamental
results, the problem of obtaining valid and efficient prediction sets has not
been thoroughly covered. Indeed, the great majority of methods currently in the
literature rely on strong distributional assumptions (e.g, Gaussianity),
dimension reduction techniques and/or asymptotic arguments. In this work, we
propose a new nonparametric approach in the field of Conformal Prediction based
on a new family of nonconformity measures inducing conformal predictors able to
create closed-form finite-sample valid or exact prediction sets under very
minimal distributional assumptions. In addition, our proposal ensures that the
prediction sets obtained are bands, an essential feature in the functional
setting that allows the visualization and interpretation of such sets. The
procedure is also fast, scalable, does not rely on functional dimension
reduction techniques and allows the user to select different nonconformity
measures depending on the problem at hand always obtaining valid bands. Within
this family of measures, we propose also a specific measure leading to
prediction bands asymptotically no less efficient than those with constant
width.
| stat |
Testing Cannot Tell Whether Ballot-Marking Devices Alter Election
Outcomes | Like all computerized systems, ballot-marking devices (BMDs) can be hacked,
misprogrammed, and misconfigured. Several approaches to testing BMDs have been
proposed. In _logic and accuracy_ (_L&A_) tests, trusted agents input known
test patterns into the BMD and check whether the printout matches. In
_parallel_ or _live_ testing, agents use the BMDs on election day, emulating
voters. In _passive_ testing, agents monitor the rate at which voters "spoil"
ballots and request another opportunity to mark a ballot: an anomalously high
rate might result from BMD malfunctions. In practice, none of these methods can
protect against outcome-altering problems. L&A testing is ineffective in part
because BMDs "know" the time and date of the test and the election. Neither L&A
nor parallel testing can probe even a small fraction of the possible voting
transactions that could comprise enough votes to change outcomes. Under mild
assumptions, to develop a model of voter interactions with BMDs accurate enough
to ensure that parallel tests could reliably detect changes to 5% of the votes
(which could change margins by 10% or more) would require monitoring the
behavior of more than a million voters in each jurisdiction in minute
detail---but the median turnout by jurisdiction in the U.S. is under 3000
voters. Given an accurate model of voter behavior, the number of tests required
is still larger than the turnout in a typical U.S. jurisdiction. Under
optimistic assumptions, passive testing that has a 99% chance of detecting a 1%
change to the margin with a 1% false alarm rate is impossible in jurisdictions
with fewer than about 1 million voters, even if the "normal" spoiled ballot
rate were known exactly and did not vary from election to election and place to
place.
| stat |
Technical Background for "A Precision Medicine Approach to Develop and
Internally Validate Optimal Exercise and Weight Loss Treatments for
Overweight and Obese Adults with Knee Osteoarthritis" | We provide additional statistical background for the methodology developed in
the clinical analysis of knee osteoarthritis in "A Precision Medicine Approach
to Develop and Internally Validate Optimal Exercise and Weight Loss Treatments
for Overweight and Obese Adults with Knee Osteoarthritis" (Jiang et al. 2020).
Jiang et al. 2020 proposed a pipeline to learn optimal treatment rules with
precision medicine models and compared them with zero-order models with a
Z-test. The model performance was based on value functions, a scalar that
predicts the future reward of each decision rule. The jackknife (i.e.,
leave-one-out cross validation) method was applied to estimate the value
function and its variance of several outcomes in IDEA. IDEA is a randomized
clinical trial studying three interventions (exercise (E), dietary weight loss
(D), and D+E) on overweight and obese participants with knee osteoarthritis. In
this report, we expand the discussion and justification with additional
statistical background. We elaborate more on the background of precision
medicine, the derivation of the jackknife estimator of value function and its
estimated variance, the consistency property of jackknife estimator, as well as
additional simulation results that reflect more of the performance of jackknife
estimators. We recommend reading Jiang et al. 2020 for clinical application and
interpretation of the optimal ITR of knee osteoarthritis as well as the overall
understanding of the pipeline and recommend using this article to understand
the underlying statistical derivation and methodology.
| stat |
Adaptive Sensor Placement for Continuous Spaces | We consider the problem of adaptively placing sensors along an interval to
detect stochastically-generated events. We present a new formulation of the
problem as a continuum-armed bandit problem with feedback in the form of
partial observations of realisations of an inhomogeneous Poisson process. We
design a solution method by combining Thompson sampling with nonparametric
inference via increasingly granular Bayesian histograms and derive an
$\tilde{O}(T^{2/3})$ bound on the Bayesian regret in $T$ rounds. This is
coupled with the design of an efficent optimisation approach to select actions
in polynomial time. In simulations we demonstrate our approach to have
substantially lower and less variable regret than competitor algorithms.
| stat |
Estimating the treatment effect of the juvenile stay-at-home order on
SARS-CoV-2 infection spread in Saline County, Arkansas | We investigate the treatment effect of the juvenile stay-at-home order
(JSAHO) adopted in Saline County, Arkansas, from April 6 to May 7, in
mitigating the growth of SARS-CoV-2 infection rates. To estimate the
counterfactual control outcome for Saline County, we apply
Difference-in-Differences and Synthetic Control design methodologies. Both
approaches show that stay-at-home order (SAHO) significantly reduced the growth
rate of the infections in Saline County during the period the policy was in
effect, contrary to some of the findings in the literature that cast doubt on
the general causal impact of SAHO with narrower scopes.
| stat |
Vertex Nomination, Consistent Estimation, and Adversarial Modification | Given a pair of graphs $G_1$ and $G_2$ and a vertex set of interest in $G_1$,
the vertex nomination (VN) problem seeks to find the corresponding vertices of
interest in $G_2$ (if they exist) and produce a rank list of the vertices in
$G_2$, with the corresponding vertices of interest in $G_2$ concentrating,
ideally, at the top of the rank list. In this paper, we define and derive the
analogue of Bayes optimality for VN with multiple vertices of interest, and we
define the notion of maximal consistency classes in vertex nomination. This
theory forms the foundation for a novel VN adversarial contamination model, and
we demonstrate with real and simulated data that there are VN schemes that
perform effectively in the uncontaminated setting, and adversarial network
contamination adversely impacts the performance of our VN scheme. We further
define a network regularization method for mitigating the impact of the
adversarial contamination, and we demonstrate the effectiveness of
regularization in both real and synthetic data.
| stat |
Prediction meets causal inference: the role of treatment in clinical
prediction models | In this paper we study approaches for dealing with treatment when developing
a clinical prediction model. Analogous to the estimand framework recently
proposed by the European Medicines Agency for clinical trials, we propose a
`predictimand' framework of different questions that may be of interest when
predicting risk in relation to treatment started after baseline. We provide a
formal definition of the estimands matching these questions, give examples of
settings in which each is useful and discuss appropriate estimators including
their assumptions. We illustrate the impact of the predictimand choice in a
dataset of patients with end-stage kidney disease. We argue that clearly
defining the estimand is equally important in prediction research as in causal
inference.
| stat |
Safe Screening Rules for $\ell_0$-Regression | We give safe screening rules to eliminate variables from regression with
$\ell_0$ regularization or cardinality constraint. These rules are based on
guarantees that a feature may or may not be selected in an optimal solution.
The screening rules can be computed from a convex relaxation solution in linear
time, without solving the $\ell_0$ optimization problem. Thus, they can be used
in a preprocessing step to safely remove variables from consideration apriori.
Numerical experiments on real and synthetic data indicate that, on average,
76\% of the variables can be fixed to their optimal values, hence, reducing the
computational burden for optimization substantially. Therefore, the proposed
fast and effective screening rules extend the scope of algorithms for
$\ell_0$-regression to larger data sets.
| stat |
Randomization Inference beyond the Sharp Null: Bounded Null Hypotheses
and Quantiles of Individual Treatment Effects | Randomization (a.k.a. permutation) inference is typically interpreted as
testing Fisher's "sharp" null hypothesis that all effects are exactly zero.
This hypothesis is often criticized as uninteresting and implausible. We show,
however, that many randomization tests are also valid for a "bounded" null
hypothesis under which effects are all negative (or positive) for all units but
otherwise heterogeneous. The bounded null is closely related to important
concepts such as monotonicity and Pareto efficiency. Inverting tests of this
hypothesis yields confidence intervals for the maximum (or minimum) individual
treatment effect. We then extend randomization tests to infer other quantiles
of individual effects, which can be used to infer the proportion of units with
effects larger (or smaller) than any threshold. The proposed confidence
intervals for all quantiles of individual effects are simultaneously valid, in
the sense that no correction due to multiple analyses is needed. In sum, we
provide a broader justification for Fisher randomization tests, and develop
exact nonparametric inference for quantiles of heterogeneous individual
effects. We illustrate our methods with simulations and applications, where we
find that Stephenson rank statistics often provide the most informative
results.
| stat |
A Methodology to Derive Global Maps of Leaf Traits Using Remote Sensing
and Climate Data | This paper introduces a modular processing chain to derive global
high-resolution maps of leaf traits. In particular, we present global maps at
500 m resolution of specific leaf area, leaf dry matter content, leaf nitrogen
and phosphorus content per dry mass, and leaf nitrogen/phosphorus ratio. The
processing chain exploits machine learning techniques along with optical remote
sensing data (MODIS/Landsat) and climate data for gap filling and up-scaling of
in-situ measured leaf traits. The chain first uses random forests regression
with surrogates to fill gaps in the database ($> 45 \% $ of missing entries)
and maximize the global representativeness of the trait dataset. Along with the
estimated global maps of leaf traits, we provide associated uncertainty
estimates derived from the regression models. The process chain is modular, and
can easily accommodate new traits, data streams (traits databases and remote
sensing data), and methods. The machine learning techniques applied allow
attribution of information gain to data input and thus provide the opportunity
to understand trait-environment relationships at the plant and ecosystem
scales.
| stat |
Model-aware Quantile Regression for Discrete Data | Quantile regression relates the quantile of the response to a linear
predictor. For a discrete response distributions, like the Poission, Binomial
and the negative Binomial, this approach is not feasible as the quantile
function is not bijective. We argue to use a continuous model-aware
interpolation of the quantile function, allowing for proper quantile inference
while retaining model interpretation. This approach allows for proper
uncertainty quantification and mitigates the issue of quantile crossing. Our
reanalysis of hospitalisation data considered in Congdon (2017) shows the
advantages of our proposal as well as introducing a novel method to exploit
quantile regression in the context of disease mapping.
| stat |
Density Estimation using Entropy Maximization for Semi-continuous Data | Semi-continuous data comes from a distribution that is a mixture of the point
mass at zero and a continuous distribution with support on the positive real
line. A clear example is the daily rainfall data. In this paper, we present a
novel algorithm to estimate the density function for semi-continuous data using
the principle of maximum entropy. Unlike existing methods in the literature,
our algorithm needs only the sample values of the constraint functions in the
entropy maximization problem and does not need the entire sample. Using
simulations, we show that the estimate of the entropy produced by our algorithm
has significantly less bias compared to existing methods. An application to the
daily rainfall data is provided.
| stat |
Assessing variable activity for Bayesian regression trees | Bayesian Additive Regression Trees (BART) are non-parametric models that can
capture complex exogenous variable effects. In any regression problem, it is
often of interest to learn which variables are most active. Variable activity
in BART is usually measured by counting the number of times a tree splits for
each variable. Such one-way counts have the advantage of fast computations.
Despite their convenience, one-way counts have several issues. They are
statistically unjustified, cannot distinguish between main effects and
interaction effects, and become inflated when measuring interaction effects. An
alternative method well-established in the literature is Sobol' indices, a
variance-based global sensitivity analysis technique. However, these indices
often require Monte Carlo integration, which can be computationally expensive.
This paper provides analytic expressions for Sobol' indices for BART posterior
samples. These expressions are easy to interpret and are computationally
feasible. Furthermore, we will show a fascinating connection between
first-order (main-effects) Sobol' indices and one-way counts. We also introduce
a novel ranking method, and use this to demonstrate that the proposed indices
preserve the Sobol'-based rank order of variable importance. Finally, we
compare these methods using analytic test functions and the En-ROADS climate
impacts simulator.
| stat |
Data-driven Analytical Models of COVID-2019 for Epidemic Prediction,
Clinical Diagnosis, Policy Effectiveness and Contact Tracing: A Survey | The widely spread CoronaVirus Disease (COVID)-19 is one of the worst
infectious disease outbreaks in history and has become an emergency of primary
international concern. As the pandemic evolves, academic communities have been
actively involved in various capacities, including accurate epidemic
estimation, fast clinical diagnosis, policy effectiveness evaluation and
development of contract tracing technologies. There are more than 23,000
academic papers on the COVID-19 outbreak, and this number is doubling every 20
days while the pandemic is still on-going [1]. The literature, however, at its
early stage, lacks a comprehensive survey from a data analytics perspective. In
this paper, we review the latest models for analyzing COVID19 related data,
conduct post-publication model evaluations and cross-model comparisons, and
collect data sources from different projects.
| stat |
Globally intensity-reweighted estimators for $K$- and pair correlation
functions | We introduce new estimators of the inhomogeneous $K$-function and the pair
correlation function of a spatial point process as well as the cross
$K$-function and the cross pair correlation function of a bivariate spatial
point process under the assumption of second-order intensity-reweighted
stationarity. These estimators rely on a 'global' normalization factor which
depends on an aggregation of the intensity function, whilst the existing
estimators depend 'locally' on the intensity function at the individual
observed points. The advantages of our new global estimators over the existing
local estimators are demonstrated by theoretical considerations and a
simulation study.
| stat |
Black-box Certification and Learning under Adversarial Perturbations | We formally study the problem of classification under adversarial
perturbations, both from the learner's perspective, and from the viewpoint of a
third-party who aims at certifying the robustness of a given black-box
classifier. We analyze a PAC-type framework of semi-supervised learning and
identify possibility and impossibility results for proper learning of
VC-classes in this setting. We further introduce and study a new setting of
black-box certification under limited query budget. We analyze this for various
classes of predictors and types of perturbation. We also consider the viewpoint
of a black-box adversary that aims at finding adversarial examples, showing
that the existence of an adversary with polynomial query complexity implies the
existence of a robust learner with small sample complexity.
| stat |
Optimal Shape Control via $L_\infty$ Loss for Composite Fuselage
Assembly | Shape control is critical to ensure the quality of composite fuselage
assembly. In current practice, the structures are adjusted to the design shape
in terms of the $\ell_2$ loss for further assembly without considering the
existing dimensional gap between two structures. Such practice has two
limitations: (1) the design shape may not be the optimal shape in terms of a
pair of incoming fuselages with different incoming dimensions; (2) the maximum
gap is the key concern during the fuselage assembly process. This paper
proposes an optimal shape control methodology via the $\ell_\infty$ loss for
composite fuselage assembly process by considering the existing dimensional gap
between the incoming pair of fuselages. Besides, due to the limitation on the
number of available actuators in practice, we face an important problem of
finding the best locations for the actuators among many potential locations,
which makes the problem a sparse estimation problem. We are the first to solve
the optimal shape control in fuselage assembly process using the $\ell_\infty$
model under the framework of sparse estimation, where we use the $\ell_1$
penalty to control the sparsity of the resulting estimator. From statistical
point of view, this can be formulated as the $\ell_\infty$ loss based linear
regression, and under some standard assumptions, such as the restricted
eigenvalue (RE) conditions, and the light tailed noise, the non-asymptotic
estimation error of the $\ell_1$ regularized $\ell_\infty$ linear model is
derived to be the order of $O(\sigma\sqrt{\frac{S\log p}{n}})$, which meets the
upper-bound in the existing literature. Compared to the current practice, the
case study shows that our proposed method significantly reduces the maximum gap
between two fuselages after shape adjustments.
| stat |
Nonparametric estimation of the first order Sobol indices with bootstrap
bandwidth | Suppose that $Y = \psi(X_1, \ldots, X_p)$, where $(X_1,\ldots, X_p)^\top$ are
random inputs, $Y$ is the output, and $\psi(\cdot)$ is an unknown link
function. The Sobol indices gauge the sensitivity of each $X$ against $Y$ by
estimating the regression curve's variability between them. In this paper, we
estimate these curves with a kernel-based method. The method allows to estimate
the first order indices when the link between the independent and dependent
variables is unknown. The kernel-based methods need a bandwidth to average the
observations. For finite samples, the cross-validation method is famous to
decide this bandwidth. However, it produces a structural bias. To remedy this,
we propose a bootstrap procedure which reconstruct the model residuals and
re-estimate the non-parametric regression curve. With the new set of curves,
the procedure corrects the bias in the Sobol index. To test the developed
method, we implemented simulated numerical examples with complex functions.
| stat |
A likelihood-based approach for multivariate categorical response
regression in high dimensions | We propose a penalized likelihood method to fit the bivariate categorical
response regression model. Our method allows practitioners to estimate which
predictors are irrelevant, which predictors only affect the marginal
distributions of the bivariate response, and which predictors affect both the
marginal distributions and log odds ratios. To compute our estimator, we
propose an efficient first order algorithm which we extend to settings where
some subjects have only one response variable measured, i.e., the
semi-supervised setting. We derive an asymptotic error bound which illustrates
the performance of our estimator in high-dimensional settings. Generalizations
to the multivariate categorical response regression model are proposed.
Finally, simulation studies and an application in pan-cancer risk prediction
demonstrate the usefulness of our method in terms of interpretability and
prediction accuracy. An R package implementing the proposed method is available
for download at github.com/ajmolstad/BvCategorical.
| stat |
Doubly robust treatment effect estimation with missing attributes | Missing attributes are ubiquitous in causal inference, as they are in most
applied statistical work. In this paper, we consider various sets of
assumptions under which causal inference is possible despite missing attributes
and discuss corresponding approaches to average treatment effect estimation,
including generalized propensity score methods and multiple imputation. Across
an extensive simulation study, we show that no single method systematically
out-performs others. We find, however, that doubly robust modifications of
standard methods for average treatment effect estimation with missing data
repeatedly perform better than their non-doubly robust baselines; for example,
doubly robust generalized propensity score methods beat inverse-weighting with
the generalized propensity score. This finding is reinforced in an analysis of
an observations study on the effect on mortality of tranexamic acid
administration among patients with traumatic brain injury in the context of
critical care management. Here, doubly robust estimators recover confidence
intervals that are consistent with evidence from randomized trials, whereas
non-doubly robust estimators do not.
| stat |
Nonparametric inference for interventional effects with multiple
mediators | Understanding the pathways whereby an intervention has an effect on an
outcome is a common scientific goal. A rich body of literature provides various
decompositions of the total intervention effect into pathway specific effects.
Interventional direct and indirect effects provide one such decomposition.
Existing estimators of these effects are based on parametric models with
confidence interval estimation facilitated via the nonparametric bootstrap. We
provide theory that allows for more flexible, possibly machine learning-based,
estimation techniques to be considered. In particular, we establish weak
convergence results that facilitate the construction of closed-form confidence
intervals and hypothesis tests. Finally, we demonstrate multiple robustness
properties of the proposed estimators. Simulations show that inference based on
large-sample theory has adequate small-sample performance. Our work thus
provides a means of leveraging modern statistical learning techniques in
estimation of interventional mediation effects.
| stat |
A Step by Step Mathematical Derivation and Tutorial on Kalman Filters | We present a step by step mathematical derivation of the Kalman filter using
two different approaches. First, we consider the orthogonal projection method
by means of vector-space optimization. Second, we derive the Kalman filter
using Bayesian optimal filtering. We provide detailed proofs for both methods
and each equation is expanded in detail.
| stat |
Tackling Algorithmic Bias in Neural-Network Classifiers using
Wasserstein-2 Regularization | The increasingly common use of neural network classifiers in industrial and
social applications of image analysis has allowed impressive progress these
last years. Such methods are however sensitive to algorithmic bias, i.e. to an
under- or an over-representation of positive predictions or to higher
prediction errors in specific subgroups of images. We then introduce in this
paper a new method to temper the algorithmic bias in Neural-Network based
classifiers. Our method is Neural-Network architecture agnostic and scales well
to massive training sets of images. It indeed only overloads the loss function
with a Wasserstein-2 based regularization term whose gradient can be computed
at a reasonable algorithmic cost. This makes it possible to use our regularised
loss with standard stochastic gradient-descent strategies. The good behavior of
our method is assessed on the Adult census, MNIST, and CelebA datasets.
| stat |
Penalized regression calibration: a method for the prediction of
survival outcomes using complex longitudinal and high-dimensional data | Longitudinal and high-dimensional measurements have become increasingly
common in biomedical research. However, methods to predict survival outcomes
using covariates that are both longitudinal and high-dimensional are currently
missing. In this article we propose penalized regression calibration (PRC), a
method that can be employed to predict survival in such situations.
PRC comprises three modelling steps: first, the trajectories described by the
longitudinal predictors are flexibly modelled through the specification of
multivariate latent process mixed models. Second, subject-specific summaries of
the longitudinal trajectories are derived from the fitted mixed effects models.
Third, the time to event outcome is predicted using the subject-specific
summaries as covariates in a penalized Cox model.
To ensure a proper internal validation of the fitted PRC models, we
furthermore develop a cluster bootstrap optimism correction procedure (CBOCP)
that allows to correct for the optimistic bias of apparent measures of
predictiveness.
After studying the behaviour of PRC via simulations, we conclude by
illustrating an application of PRC to data from an observational study that
involved patients affected by Duchenne muscular dystrophy (DMD), where the goal
is predict time to loss of ambulation using longitudinal blood biomarkers.
| stat |
mlr3proba: An R Package for Machine Learning in Survival Analysis | As machine learning has become increasingly popular over the last few
decades, so too has the number of machine learning interfaces for implementing
these models. Whilst many R libraries exist for machine learning, very few
offer extended support for survival analysis. This is problematic considering
its importance in fields like medicine, bioinformatics, economics, engineering,
and more. mlr3proba provides a comprehensive machine learning interface for
survival analysis and connects with mlr3's general model tuning and
benchmarking facilities to provide a systematic infrastructure for survival
modeling and evaluation.
| stat |
Visualizing High Dimensional Dynamical Processes | Manifold learning techniques for dynamical systems and time series have shown
their utility for a broad spectrum of applications in recent years. While these
methods are effective at learning a low-dimensional representation, they are
often insufficient for visualizing the global and local structure of the data.
In this paper, we present DIG (Dynamical Information Geometry), a visualization
method for multivariate time series data that extracts an information geometry
from a diffusion framework. Specifically, we implement a novel group of
distances in the context of diffusion operators, which may be useful to reveal
structure in the data that may not be accessible by the commonly used diffusion
distances. Finally, we present a case study applying our visualization tool to
EEG data to visualize sleep stages.
| stat |
Nowcasting COVID-19 incidence indicators during the Italian first
outbreak | A novel parametric regression model is proposed to fit incidence data
typically collected during epidemics. The proposal is motivated by real-time
monitoring and short-term forecasting of the main epidemiological indicators
within the first outbreak of COVID-19 in Italy. Accurate short-term
predictions, including the potential effect of exogenous or external variables
are provided; this ensures to accurately predict important characteristics of
the epidemic (e.g., peak time and height), allowing for a better allocation of
health resources over time. Parameters estimation is carried out in a maximum
likelihood framework. All computational details required to reproduce the
approach and replicate the results are provided.
| stat |
Efficient Bayesian reduced rank regression using Langevin Monte Carlo
approach | The problem of Bayesian reduced rank regression is considered in this paper.
We propose, for the first time, to use Langevin Monte Carlo method in this
problem. A spectral scaled Student prior distrbution is used to exploit the
underlying low-rank structure of the coefficient matrix. We show that our
algorithms are significantly faster than the Gibbs sampler in high-dimensional
setting. Simulation results show that our proposed algorithms for Bayesian
reduced rank regression are comparable to the state-of-the-art method where the
rank is chosen by cross validation.
| stat |
Fast calculation of p-values for one-sided Kolmogorov-Smirnov type
statistics | We present a method for computing exact p-values for a large family of
one-sided continuous goodness-of-fit statistics. This includes the higher
criticism statistic, one-sided weighted Kolmogorov-Smirnov statistics, and the
one-sided Berk-Jones statistics. For a sample size of 10,000, our method takes
merely 0.15 seconds to run and it scales to sample sizes in the hundreds of
thousands. This allows practitioners working on genome-wide association studies
and other high-dimensional analyses to use exact finite-sample computations
instead of statistic-specific approximation schemes.
Our work has other applications in statistics, including power analysis,
finding alpha-level thresholds for goodness-of-fit tests, and the construction
of confidence bands for the empirical distribution function. The algorithm is
based on a reduction to the boundary-crossing probability of a pure jump
process and is also applicable to fields outside of statistics, for example in
financial risk modeling.
| stat |
An enriched mixture model for functional clustering | There is an increasingly rich literature about Bayesian nonparametric models
for clustering functional observations. However, most of the recent proposals
rely on infinite-dimensional characterizations that might lead to overly
complex cluster solutions. In addition, while prior knowledge about the
functional shapes is typically available, its practical exploitation might be a
difficult modeling task. Motivated by an application in e-commerce, we propose
a novel enriched Dirichlet mixture model for functional data. Our proposal
accommodates the incorporation of functional constraints while bounding the
model complexity. To clarify the underlying partition mechanism, we
characterize the prior process through a P\'olya urn scheme. These features
lead to a very interpretable clustering method compared to available
techniques. To overcome computational bottlenecks, we employ a variational
Bayes approximation for tractable posterior inference.
| stat |
Bayesian Meta-analysis of Rare Events with Non-ignorable Missing Data | Meta-analysis is a powerful tool for drug safety assessment by synthesizing
treatment-related toxicological findings from independent clinical trials.
However, published clinical studies may or may not report all adverse events
(AEs) if the observed number of AEs were fewer than a pre-specified
study-dependent cutoff. Subsequently, with censored information ignored, the
estimated incidence rate of AEs could be significantly biased. To address this
non-ignorable missing data problem in meta-analysis, we propose a Bayesian
multilevel regression model to accommodate the censored rare event data. The
performance of the proposed Bayesian model of censored data compared to other
existing methods is demonstrated through simulation studies under various
censoring scenarios. Finally, the proposed approach is illustrated using data
from a recent meta-analysis of 125 clinical trials involving PD-1/PD-L1
inhibitors with respect to their toxicity profiles.
| stat |
Greedy clustering of count data through a mixture of multinomial PCA | Count data is becoming more and more ubiquitous in a wide range of
applications, with datasets growing both in size and in dimension. In this
context, an increasing amount of work is dedicated to the construction of
statistical models directly accounting for the discrete nature of the data.
Moreover, it has been shown that integrating dimension reduction to clustering
can drastically improve performance and stability. In this paper, we rely on
the mixture of multinomial PCA, a mixture model for the clustering of count
data, also known as the probabilistic clustering-projection model in the
literature. Related to the latent Dirichlet allocation model, it offers the
flexibility of topic modeling while being able to assign each observation to a
unique cluster. We introduce a greedy clustering algorithm, where inference and
clustering are jointly done by mixing a classification variational expectation
maximization algorithm, with a branch & bound like strategy on a variational
lower bound. An integrated classification likelihood criterion is derived for
model selection, and a thorough study with numerical experiments is proposed to
assess both the performance and robustness of the method. Finally, we
illustrate the qualitative interest of the latter in a real-world application,
for the clustering of anatomopathological medical reports, in partnership with
expert practitioners from the Institut Curie hospital.
| stat |
Calibration of inexact computer models with heteroscedastic errors | Computer models are commonly used to represent a wide range of real systems,
but they often involve some unknown parameters. Estimating the parameters by
collecting physical data becomes essential in many scientific fields, ranging
from engineering to biology. However, most of the existing methods are
developed under the assumption that the physical data contains homoscedastic
measurement errors. Motivated by an experiment of plant relative growth rates
where replicates are available, we propose a new calibration method for inexact
computer models with heteroscedastic measurement errors. Asymptotic properties
of the parameter estimators are derived, and a goodness-of-fit test is
developed to detect the presence of heteroscedasticity. Numerical examples and
empirical studies demonstrate that the proposed method not only yields accurate
parameter estimation, but it also provides accurate predictions for physical
data in the presence of both heteroscedasticity and model misspecification.
| stat |
Estimating Graphlet Statistics via Lifting | Exploratory analysis over network data is often limited by the ability to
efficiently calculate graph statistics, which can provide a model-free
understanding of the macroscopic properties of a network. We introduce a
framework for estimating the graphlet count---the number of occurrences of a
small subgraph motif (e.g. a wedge or a triangle) in the network. For massive
graphs, where accessing the whole graph is not possible, the only viable
algorithms are those that make a limited number of vertex neighborhood queries.
We introduce a Monte Carlo sampling technique for graphlet counts, called {\em
Lifting}, which can simultaneously sample all graphlets of size up to $k$
vertices for arbitrary $k$. This is the first graphlet sampling method that can
provably sample every graphlet with positive probability and can sample
graphlets of arbitrary size $k$. We outline variants of lifted graphlet counts,
including the ordered, unordered, and shotgun estimators, random walk starts,
and parallel vertex starts. We prove that our graphlet count updates are
unbiased for the true graphlet count and have a controlled variance for all
graphlets. We compare the experimental performance of lifted graphlet counts to
the state-of-the art graphlet sampling procedures: Waddling and the pairwise
subgraph random walk.
| stat |
Modeling partitions of individuals | Despite the central role of self-assembled groups in animal and human
societies, statistical tools to explain their composition are limited. We
introduce a statistical framework for cross-sectional observations of groups
with exclusive membership to illuminate the social and organizational
mechanisms that bring people together. Drawing from stochastic models for
networks and partitions, the proposed framework introduces an exponential
family of distributions for partitions. We derive its main mathematical
properties and suggest strategies to specify and estimate such models. A case
study on hackathon events applies the developed framework to the study of
mechanisms underlying the formation of self-assembled project teams.
| stat |
Correcting a Nonparametric Two-sample Graph Hypothesis Test for Graphs
with Different Numbers of Vertices | Random graphs are statistical models that have many applications, ranging
from neuroscience to social network analysis. Of particular interest in some
applications is the problem of testing two random graphs for equality of
generating distributions. Tang et al. (2017) propose a test for this setting.
This test consists of embedding the graph into a low-dimensional space via the
adjacency spectral embedding (ASE) and subsequently using a kernel two-sample
test based on the maximum mean discrepancy. However, if the two graphs being
compared have an unequal number of vertices, the test of Tang et al. (2017) may
not be valid. We demonstrate the intuition behind this invalidity and propose a
correction that makes any subsequent kernel- or distance-based test valid. Our
method relies on sampling based on the asymptotic distribution for the ASE. We
call these altered embeddings the corrected adjacency spectral embeddings
(CASE). We show that CASE remedies the exchangeability problem of the original
test and demonstrate the validity and consistency of the test that uses CASE
via a simulation study. We also apply our proposed test to the problem of
determining equivalence of generating distributions of subgraphs in Drosophila
connectome.
| stat |
Simultaneous semi-parametric estimation of clustering and regression | We investigate the parameter estimation of regression models with fixed group
effects, when the group variable is missing while group related variables are
available. This problem involves clustering to infer the missing group variable
based on the group related variables, and regression to build a model on the
target variable given the group and eventually additional variables. Thus, this
problem can be formulated as the joint distribution modeling of the target and
of the group related variables. The usual parameter estimation strategy for
this joint model is a two-step approach starting by learning the group variable
(clustering step) and then plugging in its estimator for fitting the regression
model (regression step). However, this approach is suboptimal (providing in
particular biased regression estimates) since it does not make use of the
target variable for clustering. Thus, we claim for a simultaneous estimation
approach of both clustering and regression, in a semi-parametric framework.
Numerical experiments illustrate the benefits of our proposition by considering
wide ranges of distributions and regression models. The relevance of our new
method is illustrated on real data dealing with problems associated with high
blood pressure prevention.
| stat |
A New Causal Approach to Account for Treatment Switching in Randomized
Experiments under a Structural Cumulative Survival Model | Treatment switching in a randomized controlled trial is said to occur when a
patient randomized to one treatment arm switches to another treatment arm
during follow-up. This can occur at the point of disease progression, whereby
patients in the control arm may be offered the experimental treatment. It is
widely known that failure to account for treatment switching can seriously
dilute the estimated effect of treatment on overall survival. In this paper, we
aim to account for the potential impact of treatment switching in a re-analysis
evaluating the treatment effect of NucleosideReverse Transcriptase Inhibitors
(NRTIs) on a safety outcome (time to first severe or worse sign or symptom) in
participants receiving a new antiretroviral regimen that either included or
omitted NRTIs in the Optimized Treatment That Includes or OmitsNRTIs (OPTIONS)
trial. We propose an estimator of a treatment causal effect under a structural
cumulative survival model (SCSM) that leverages randomization as an
instrumental variable to account for selective treatment switching. Unlike
Robins' accelerated failure time model often used to address treatment
switching, the proposed approach avoids the need for artificial censoring for
estimation. We establish that the proposed estimator is uniformly consistent
and asymptotically Gaussian under standard regularity conditions. A consistent
variance estimator is also given and a simple resampling approach provides
uniform confidence bands for the causal difference comparing treatment groups
overtime on the cumulative intensity scale. We develop an R package named
"ivsacim" implementing all proposed methods, freely available to download from
R CRAN. We examine the finite performance of the estimator via extensive
simulations.
| stat |
High-Dimensional Independence Testing and Maximum Marginal Correlation | A number of universally consistent dependence measures have been recently
proposed for testing independence, such as distance correlation, kernel
correlation, multiscale graph correlation, etc. They provide a satisfactory
solution for dependence testing in low-dimensions, but often exhibit decreasing
power for high-dimensional data, a phenomenon that has been recognized but
remains mostly unchartered. In this paper, we aim to better understand the
high-dimensional testing scenarios and explore a procedure that is robust
against increasing dimension. To that end, we propose the maximum marginal
correlation method and characterize high-dimensional dependence structures via
the notion of dependent dimensions. We prove that the maximum method can be
valid and universally consistent for testing high-dimensional dependence under
regularity conditions, and demonstrate when and how the maximum method may
outperform other methods. The methodology can be implemented by most existing
dependence measures, has a superior testing power in a variety of common
high-dimensional settings, and is computationally efficient for big data
analysis when using the distance correlation chi-square test.
| stat |
Equitable and Optimal Transport with Multiple Agents | We introduce an extension of the Optimal Transport problem when multiple
costs are involved. Considering each cost as an agent, we aim to share equally
between agents the work of transporting one distribution to another. To do so,
we minimize the transportation cost of the agent who works the most. Another
point of view is when the goal is to partition equitably goods between agents
according to their heterogeneous preferences. Here we aim to maximize the
utility of the least advantaged agent. This is a fair division problem. Like
Optimal Transport, the problem can be cast as a linear optimization problem.
When there is only one agent, we recover the Optimal Transport problem. When
two agents are considered, we are able to recover Integral Probability Metrics
defined by $\alpha$-H\"older functions, which include the widely-known Dudley
metric. To the best of our knowledge, this is the first time a link is given
between the Dudley metric and Optimal Transport. We provide an entropic
regularization of that problem which leads to an alternative algorithm faster
than the standard linear program.
| stat |
The Theory Behind Overfitting, Cross Validation, Regularization,
Bagging, and Boosting: Tutorial | In this tutorial paper, we first define mean squared error, variance,
covariance, and bias of both random variables and classification/predictor
models. Then, we formulate the true and generalization errors of the model for
both training and validation/test instances where we make use of the Stein's
Unbiased Risk Estimator (SURE). We define overfitting, underfitting, and
generalization using the obtained true and generalization errors. We introduce
cross validation and two well-known examples which are $K$-fold and
leave-one-out cross validations. We briefly introduce generalized cross
validation and then move on to regularization where we use the SURE again. We
work on both $\ell_2$ and $\ell_1$ norm regularizations. Then, we show that
bootstrap aggregating (bagging) reduces the variance of estimation. Boosting,
specifically AdaBoost, is introduced and it is explained as both an additive
model and a maximum margin model, i.e., Support Vector Machine (SVM). The upper
bound on the generalization error of boosting is also provided to show why
boosting prevents from overfitting. As examples of regularization, the theory
of ridge and lasso regressions, weight decay, noise injection to input/weights,
and early stopping are explained. Random forest, dropout, histogram of oriented
gradients, and single shot multi-box detector are explained as examples of
bagging in machine learning and computer vision. Finally, boosting tree and SVM
models are mentioned as examples of boosting.
| stat |
Multi-resolution Multi-task Gaussian Processes | We consider evidence integration from potentially dependent observation
processes under varying spatio-temporal sampling resolutions and noise levels.
We develop a multi-resolution multi-task (MRGP) framework while allowing for
both inter-task and intra-task multi-resolution and multi-fidelity. We develop
shallow Gaussian Process (GP) mixtures that approximate the difficult to
estimate joint likelihood with a composite one and deep GP constructions that
naturally handle biases in the mean. By doing so, we generalize and outperform
state of the art GP compositions and offer information-theoretic corrections
and efficient variational approximations. We demonstrate the competitiveness of
MRGPs on synthetic settings and on the challenging problem of hyper-local
estimation of air pollution levels across London from multiple sensing
modalities operating at disparate spatio-temporal resolutions.
| stat |
A New Framework for Distance and Kernel-based Metrics in High Dimensions | The paper presents new metrics to quantify and test for (i) the equality of
distributions and (ii) the independence between two high-dimensional random
vectors. We show that the energy distance based on the usual Euclidean distance
cannot completely characterize the homogeneity of two high-dimensional
distributions in the sense that it only detects the equality of means and the
traces of covariance matrices in the high-dimensional setup. We propose a new
class of metrics which inherits the desirable properties of the energy distance
and maximum mean discrepancy/(generalized) distance covariance and the
Hilbert-Schmidt Independence Criterion in the low-dimensional setting and is
capable of detecting the homogeneity of/completely characterizing independence
between the low-dimensional marginal distributions in the high dimensional
setup. We further propose t-tests based on the new metrics to perform
high-dimensional two-sample testing/independence testing and study their
asymptotic behavior under both high dimension low sample size (HDLSS) and high
dimension medium sample size (HDMSS) setups. The computational complexity of
the t-tests only grows linearly with the dimension and thus is scalable to very
high dimensional data. We demonstrate the superior power behavior of the
proposed tests for homogeneity of distributions and independence via both
simulated and real datasets.
| stat |
Efficient nonparametric statistical inference on population feature
importance using Shapley values | The true population-level importance of a variable in a prediction task
provides useful knowledge about the underlying data-generating mechanism and
can help in deciding which measurements to collect in subsequent experiments.
Valid statistical inference on this importance is a key component in
understanding the population of interest. We present a computationally
efficient procedure for estimating and obtaining valid statistical inference on
the Shapley Population Variable Importance Measure (SPVIM). Although the
computational complexity of the true SPVIM scales exponentially with the number
of variables, we propose an estimator based on randomly sampling only
$\Theta(n)$ feature subsets given $n$ observations. We prove that our estimator
converges at an asymptotically optimal rate. Moreover, by deriving the
asymptotic distribution of our estimator, we construct valid confidence
intervals and hypothesis tests. Our procedure has good finite-sample
performance in simulations, and for an in-hospital mortality prediction task
produces similar variable importance estimates when different machine learning
algorithms are applied.
| stat |
Conditional Selective Inference for Robust Regression and Outlier
Detection using Piecewise-Linear Homotopy Continuation | In practical data analysis under noisy environment, it is common to first use
robust methods to identify outliers, and then to conduct further analysis after
removing the outliers. In this paper, we consider statistical inference of the
model estimated after outliers are removed, which can be interpreted as a
selective inference (SI) problem. To use conditional SI framework, it is
necessary to characterize the events of how the robust method identifies
outliers. Unfortunately, the existing methods cannot be directly used here
because they are applicable to the case where the selection events can be
represented by linear/quadratic constraints. In this paper, we propose a
conditional SI method for popular robust regressions by using homotopy method.
We show that the proposed conditional SI method is applicable to a wide class
of robust regression and outlier detection methods and has good empirical
performance on both synthetic data and real data experiments.
| stat |
Adaptive Bayesian Changepoint Analysis and Local Outlier Scoring | We introduce global-local shrinkage priors into a Bayesian dynamic linear
model to adaptively estimate both changepoints and local outliers in a novel
model we call Adaptive Bayesian Changepoints with Outliers (ABCO). We utilize a
state-space approach to identify a dynamic signal in the presence of outliers
and measurement error with stochastic volatility. We find that global state
equation parameters are inadequate for most real applications and we include
local parameters to track noise at each time-step. This setup provides a
flexible framework to detect unspecified changepoints in complex series, such
as those with large interruptions in local trends, with robustness to outliers
and heteroskedastic noise. ABCO may also be used as a robust Bayesian trend
filter that can reconstruct interrupted time series. We detail the extension of
our approach to time-varying parameter estimation within dynamic regression
analysis to identify structural breaks. Finally, we compare our algorithm
against several alternatives to demonstrate its efficacy in diverse simulation
scenarios and two empirical examples.
| stat |
Testing for arbitrary interference on experimentation platforms | Experimentation platforms are essential to modern large technology companies,
as they are used to carry out many randomized experiments daily. The classic
assumption of no interference among users, under which the outcome of one user
does not depend on the treatment assigned to other users, is rarely tenable on
such platforms. Here, we introduce an experimental design strategy for testing
whether this assumption holds. Our approach is in the spirit of the
Durbin-Wu-Hausman test for endogeneity in econometrics, where multiple
estimators return the same estimate if and only if the null hypothesis holds.
The design that we introduce makes no assumptions on the interference model
between units, nor on the network among the units, and has a sharp bound on the
variance and an implied analytical bound on the type I error rate. We discuss
how to apply the proposed design strategy to large experimentation platforms,
and we illustrate it in the context of an experiment on the LinkedIn platform.
| stat |
Robust Approximate Bayesian Computation: An Adjustment Approach | We propose a novel approach to approximate Bayesian computation (ABC) that
seeks to cater for possible misspecification of the assumed model. This new
approach can be equally applied to rejection-based ABC and to popular
regression adjustment ABC. We demonstrate that this new approach mitigates the
poor performance of regression adjusted ABC that can eventuate when the model
is misspecified. In addition, this new adjustment approach allows us to detect
which features of the observed data can not be reliably reproduced by the
assumed model. A series of simulated and empirical examples illustrate this new
approach.
| stat |
A Registration-free approach for Statistical Process Control of 3D
scanned objects via FEM | Recent work in on-line Statistical Process Control (SPC) of manufactured
3-dimensional (3-D) objects has been proposed based on the estimation of the
spectrum of the Laplace-Beltrami (LB) operator, a differential operator that
encodes the geometrical features of a manifold and is widely used in Machine
Learning (i.e., Manifold Learning). The resulting spectra are an intrinsic
geometrical feature of each part, and thus can be compared between parts
avoiding the part to part registration (or "part localization") pre-processing
or the need for equal size meshes, characteristics which are required in
previous approaches for SPC of 3D parts. The recent spectral SPC methods,
however, are limited to monitoring surface data from objects such that the
scanned meshes have no boundaries, holes, or missing portions. In this paper,
we extend spectral methods by first considering a more accurate and general
estimator of the LB spectrum that is obtained by application of Finite Element
Methods (FEM) to the solution of Helmholtz's equation with boundaries. It is
shown how the new spectral FEM approach, while it retains the advantages of not
requiring part localization/registration or equal size datasets scanned from
each part, it provides more accurate spectrum estimates, which results in
faster detection of out of control conditions than earlier methods, can be
applied to both mesh or volumetric (solid) scans, and furthermore, it is shown
how it can be applied to partial scans that result in open meshes (surface or
volumetric) with boundaries, increasing the practical applicability of the
methods. The present work brings SPC methods closer to contemporary research in
Computer Graphics and Manifold Learning. MATLAB code that reproduces the
examples of this paper is provided in the supplementary materials.
| stat |
Quantifying treatment differences in confirmatory trials under
non-proportional hazards | Proportional hazards are a common assumption when designing confirmatory
clinical trials in oncology. With the emergence of immunotherapy and novel
targeted therapies, departure from the proportional hazard assumption is not
rare in nowadays clinical research. Under non-proportional hazards, the hazard
ratio does not have a straightforward clinical interpretation, and the log-rank
test is no longer the most powerful statistical test even though it is still
valid. Nevertheless, the log-rank test and the hazard ratio are still the
primary analysis tools, and traditional approaches such as sample size increase
are still proposed to account for the impact of non-proportional hazards. The
weighed log-rank test and the test based on the restricted mean survival time
(RMST) are receiving a lot of attention as a potential alternative to the
log-rank test. We conduct a simulation study comparing the performance and
operating characteristics of the log-rank test, the weighted log-rank test and
the test based on the RMST, including a treatment effect estimation, under
different non-proportional hazards patterns. Results show that, under
non-proportional hazards, the hazard ratio and weighted hazard ratio have no
straightforward clinical interpretation whereas the RMST ratio can be
interpreted regardless of the proportional hazards assumption. In terms of
power, the RMST achieves a similar performance when compared to the log-rank
test.
| stat |
A Comprehensive Survey of Inverse Uncertainty Quantification of Physical
Model Parameters in Nuclear System Thermal-Hydraulics Codes | Uncertainty Quantification (UQ) is an essential step in computational model
validation because assessment of the model accuracy requires a concrete,
quantifiable measure of uncertainty in the model predictions. The concept of UQ
in the nuclear community generally means forward UQ (FUQ), in which the
information flow is from the inputs to the outputs. Inverse UQ (IUQ), in which
the information flow is from the model outputs and experimental data to the
inputs, is an equally important component of UQ but has been significantly
underrated until recently. FUQ requires knowledge in the input uncertainties
which has been specified by expert opinion or user self-evaluation. IUQ is
defined as the process to inversely quantify the input uncertainties based on
experimental data. This review paper aims to provide a comprehensive and
comparative discussion of the major aspects of the IUQ methodologies that have
been used on the physical models in system thermal-hydraulics codes. IUQ
methods can be categorized by three main groups: frequentist (deterministic),
Bayesian (probabilistic), and empirical (design-of-experiments). We used eight
metrics to evaluate an IUQ method, including solidity, complexity,
accessibility, independence, flexibility, comprehensiveness, transparency, and
tractability. Twelve IUQ methods are reviewed, compared, and evaluated based on
these eight metrics. Such comparative evaluation will provide a good guidance
for users to select a proper IUQ method based on the IUQ problem under
investigation.
| stat |
Adaptive preferential sampling in phylodynamics | Longitudinal molecular data of rapidly evolving viruses and pathogens provide
information about disease spread and complement traditional surveillance
approaches based on case count data. The coalescent is used to model the
genealogy that represents the sample ancestral relationships. The basic
assumption is that coalescent events occur at a rate inversely proportional to
the effective population size $N_{e}(t)$, a time-varying measure of genetic
diversity. When the sampling process (collection of samples over time) depends
on $N_{e}(t)$, the coalescent and the sampling processes can be jointly modeled
to improve estimation of $N_{e}(t)$. Failing to do so can lead to bias due to
model misspecification. However, the way that the sampling process depends on
the effective population size may vary over time. We introduce an approach
where the sampling process is modeled as an inhomogeneous Poisson process with
rate equal to the product of $N_{e}(t)$ and a time-varying coefficient, making
minimal assumptions on their functional shapes via Markov random field priors.
We provide scalable algorithms for inference, show the model performance
vis-a-vis alternative methods in a simulation study, and apply our model to
SARS-CoV-2 sequences from Los Angeles and Santa Clara counties. The methodology
is implemented and available in the R package adapref.
| stat |
New statistical model for misreported data with application to current
public health challenges | The main goal of this work is to present a new model able to deal with
potentially misreported continuous time series. The proposed model is able to
handle the autocorrelation structure in continuous time series data, which
might be partially or totally underreported or overreported. Its performance is
illustrated through a comprehensive simulation study considering several
autocorrelation structures and two real data applications on human
papillomavirus incidence in Girona (Catalunya, Spain) and COVID-19 incidence in
the Chinese region of Heilongjiang.
| stat |
Improved assessment of the accuracy of record linkage via an extended
MaCSim approach | Record linkage is the process of bringing together the same entity from
overlapping data sources while removing duplicates. Huge amounts of data are
now being collected by public or private organizations as well as by
researchers and individuals. Linking and analysing relevant information from
this massive data reservoir can provide new insights into society. However,
this increase in the amount of data may also increase the likelihood of
incorrectly linked records among databases. It has become increasingly
important to have effective and efficient methods for linking data from
different sources. Therefore, it becomes necessary to assess the ability of a
linking method to achieve high accuracy or to compare between methods with
respect to accuracy. In this paper, we improve on a Markov Chain based Monte
Carlo simulation approach (MaCSim) for assessing a linking method. MaCSim
utilizes two linked files that have been previously linked on similar types of
data to create an agreement matrix and then simulates the matrix using a
proposed algorithm developed to generate re-sampled versions of the agreement
matrix. A defined linking method is used in each simulation to link the files
and the accuracy of the linking method is assessed. The improvement proposed
here involves calculation of a similarity weight for every linking variable
value for each record pair, which allows partial agreement of the linking
variable values. A threshold is calculated for every linking variable based on
adjustable parameter "tolerance" for that variable. To assess the accuracy of
linking method, correctly linked proportions are investigated for each record.
The extended MaCSim approach is illustrated using a synthetic dataset provided
by the Australian Bureau of Statistics (ABS) based on realistic data settings.
Test results show higher accuracy of the assessment of linkages.
| stat |
Group design in group testing for COVID-19 : A French case-study | Group testing is a screening strategy that involves dividing a population
into several disjointed groups of subjects. In its simplest implementation,
each group is tested with a single test in the first phase, while in the second
phase only subjects in positive groups, if any, need to be tested again
individually. In this paper, we address the problem of group testing design,
which aims to determine a partition into groups of a finite population in such
a way that cardinality constraints on the size of each group and a constraint
on the expected total number of tests are satisfied while minimizing a linear
combination of the expected number of false negative and false positive
classifications. First, we show that the properties and model introduced by
Aprahmian et al. can be extended to the group test design problem, which is
then modeled as a constrained shortest path problem on a specific graph. We
design and implement an ad hoc algorithm to solve this problem. On instances
based on Sant\'e Publique France data on Covid-19 screening tests, the results
of the computational experiments are very promising.
| stat |
MLE of Jointly Constrained Mean-Covariance of Multivariate Normal
Distributions | Estimating the unconstrained mean and covariance matrix is a popular topic in
statistics. However, estimation of the parameters of $N_p(\mu,\Sigma)$ under
joint constraints such as $\Sigma\mu = \mu$ has not received much attention. It
can be viewed as a multivariate counterpart of the classical estimation problem
in the $N(\theta,\theta^2)$ distribution. In addition to the usual inference
challenges under such non-linear constraints among the parameters (curved
exponential family), one has to deal with the basic requirements of symmetry
and positive definiteness when estimating a covariance matrix. We derive the
non-linear likelihood equations for the constrained maximum likelihood
estimator of $(\mu,\Sigma)$ and solve them using iterative methods. Generally,
the MLE of covariance matrices computed using iterative methods do not satisfy
the constraints. We propose a novel algorithm to modify such (infeasible)
estimators or any other (reasonable) estimator. The key step is to re-align the
mean vector along the eigenvectors of the covariance matrix using the idea of
regression. In using the Lagrangian function for constrained MLE (Aitchison et
al. 1958), the Lagrange multiplier entangles with the parameters of interest
and presents another computational challenge. We handle this by either
iterative or explicit calculation of the Lagrange multiplier. The existence and
nature of location of the constrained MLE are explored within a data-dependent
convex set using recent results from random matrix theory. A simulation study
illustrates our methodology and shows that the modified estimators perform
better than the initial estimators from the iterative methods.
| stat |
An Algorithmic Approach to Forecasting Rare Violent Events: An
Illustration Based in IPV Perpetration | Mass violence, almost no matter how defined, is (thankfully) rare. Rare
events are very difficult to study in a systematic manner. Standard statistical
procedures can fail badly and usefully accurate forecasts of rare events often
are little more than an aspiration. We offer an unconventional approach for the
statistical analysis of rare events illustrated by an extensive case study. We
report research whose goal is to learn about the attributes of very high risk
IPV perpetrators and the circumstances associated with their IPV incidents
reported to the police. Very high risk is defined as having a high probability
of committing a repeat IPV assault in which the victim is injured. Such
individuals represent a very small fraction of all IPV perpetrators; these acts
of violence are relatively rare. To learn about them nevertheless, we apply in
a novel fashion three algorithms sequentially to data collected from a large
metropolitan police department: stochastic gradient boosting, a genetic
algorithm inspired by natural selection, and agglomerative clustering. We try
to characterize not just perpetrators who on balance are predicted to
re-offend, but who are very likely to re-offend in a manner that leads to
victim injuries. With this strategy, we learn a lot. We also provide a new way
to estimate the importance of risk predictors. There are lessons for the study
of other rare forms of violence especially when instructive forecasts are
sought. In the absence of sufficiently accurate forecasts, scarce prevention
resources cannot be allocated where they are most needed.
| stat |
Experimental Design in Two-Sided Platforms: An Analysis of Bias | We develop an analytical framework to study experimental design in two-sided
marketplaces. Many of these experiments exhibit interference, where an
intervention applied to one market participant influences the behavior of
another participant. This interference leads to biased estimates of the
treatment effect of the intervention. We develop a stochastic market model and
associated mean field limit to capture dynamics in such experiments, and use
our model to investigate how the performance of different designs and
estimators is affected by marketplace interference effects. Platforms typically
use two common experimental designs: demand-side ("customer") randomization
(CR) and supply-side ("listing") randomization (LR), along with their
associated estimators. We show that good experimental design depends on market
balance: in highly demand-constrained markets, CR is unbiased, while LR is
biased; conversely, in highly supply-constrained markets, LR is unbiased, while
CR is biased. We also introduce and study a novel experimental design based on
two-sided randomization (TSR) where both customers and listings are randomized
to treatment and control. We show that appropriate choices of TSR designs can
be unbiased in both extremes of market balance, while yielding relatively low
bias in intermediate regimes of market balance.
| stat |
Towards forecast techniques for business analysts of large commercial
data sets using matrix factorization methods | This research article suggests that there are significant benefits in
exposing demand planners to forecasting methods using matrix completion
techniques. This study aims to contribute to a better understanding of the
field of forecasting with multivariate time series prediction by focusing on
the dimension of large commercial data sets with hierarchies. This research
highlights that there has neither been sufficient academic research in this
sub-field nor dissemination among practitioners in the business sector. This
study seeks to innovate by presenting a matrix completion method for short-term
demand forecast of time series data on relevant commercial problems. Albeit
computing intensive, this method outperforms the state of the art while
remaining accessible to business users. The object of research is matrix
completion for time series in a big data context within the industry. The
subject of the research is forecasting product demand using techniques for
multivariate hierarchical time series prediction that are both precise and
accessible to non-technical business experts. Apart from a methodological
innovation, this research seeks to introduce practitioners to novel methods for
hierarchical multivariate time series prediction. The research outcome is of
interest for organizations requiring precise forecasts yet lacking the
appropriate human capital to develop them.
| stat |
Two-stage Best-scored Random Forest for Large-scale Regression | We propose a novel method designed for large-scale regression problems,
namely the two-stage best-scored random forest (TBRF). "Best-scored" means to
select one regression tree with the best empirical performance out of a certain
number of purely random regression tree candidates, and "two-stage" means to
divide the original random tree splitting procedure into two: In stage one, the
feature space is partitioned into non-overlapping cells; in stage two, child
trees grow separately on these cells. The strengths of this algorithm can be
summarized as follows: First of all, the pure randomness in TBRF leads to the
almost optimal learning rates, and also makes ensemble learning possible, which
resolves the boundary discontinuities long plaguing the existing algorithms.
Secondly, the two-stage procedure paves the way for parallel computing, leading
to computational efficiency. Last but not least, TBRF can serve as an inclusive
framework where different mainstream regression strategies such as linear
predictor and least squares support vector machines (LS-SVMs) can also be
incorporated as value assignment approaches on leaves of the child trees,
depending on the characteristics of the underlying data sets. Numerical
assessments on comparisons with other state-of-the-art methods on several
large-scale real data sets validate the promising prediction accuracy and high
computational efficiency of our algorithm.
| stat |
Spike-and-Slab Group Lassos for Grouped Regression and Sparse
Generalized Additive Models | We introduce the spike-and-slab group lasso (SSGL) for Bayesian estimation
and variable selection in linear regression with grouped variables. We further
extend the SSGL to sparse generalized additive models (GAMs), thereby
introducing the first nonparametric variant of the spike-and-slab lasso
methodology. Our model simultaneously performs group selection and estimation,
while our fully Bayes treatment of the mixture proportion allows for model
complexity control and automatic self-adaptivity to different levels of
sparsity. We develop theory to uniquely characterize the global posterior mode
under the SSGL and introduce a highly efficient block coordinate ascent
algorithm for maximum a posteriori (MAP) estimation. We further employ
de-biasing methods to provide uncertainty quantification of our estimates.
Thus, implementation of our model avoids the computational intensiveness of
Markov chain Monte Carlo (MCMC) in high dimensions. We derive posterior
concentration rates for both grouped linear regression and sparse GAMs when the
number of covariates grows at nearly exponential rate with sample size.
Finally, we illustrate our methodology through extensive simulations and data
analysis.
| stat |
Aggregated Pairwise Classification of Statistical Shapes | The classification of shapes is of great interest in diverse areas ranging
from medical imaging to computer vision and beyond. While many statistical
frameworks have been developed for the classification problem, most are
strongly tied to early formulations of the problem - with an object to be
classified described as a vector in a relatively low-dimensional Euclidean
space. Statistical shape data have two main properties that suggest a need for
a novel approach: (i) shapes are inherently infinite dimensional with strong
dependence among the positions of nearby points, and (ii) shape space is not
Euclidean, but is fundamentally curved. To accommodate these features of the
data, we work with the square-root velocity function of the curves to provide a
useful formal description of the shape, pass to tangent spaces of the manifold
of shapes at different projection points which effectively separate shapes for
pairwise classification in the training data, and use principal components
within these tangent spaces to reduce dimensionality. We illustrate the impact
of the projection point and choice of subspace on the misclassification rate
with a novel method of combining pairwise classifiers.
| stat |
Learning from Label Proportions: A Mutual Contamination Framework | Learning from label proportions (LLP) is a weakly supervised setting for
classification in which unlabeled training instances are grouped into bags, and
each bag is annotated with the proportion of each class occurring in that bag.
Prior work on LLP has yet to establish a consistent learning procedure, nor
does there exist a theoretically justified, general purpose training criterion.
In this work we address these two issues by posing LLP in terms of mutual
contamination models (MCMs), which have recently been applied successfully to
study various other weak supervision settings. In the process, we establish
several novel technical results for MCMs, including unbiased losses and
generalization error bounds under non-iid sampling plans. We also point out the
limitations of a common experimental setting for LLP, and propose a new one
based on our MCM framework.
| stat |
The $r$-value: evaluating stability with respect to distributional
shifts | Common statistical measures of uncertainty like $p$-values and confidence
intervals quantify the uncertainty due to sampling, that is, the uncertainty
due to not observing the full population. In practice, populations change
between locations and across time. This makes it difficult to gather knowledge
that transfers across data sets. We propose a measure of uncertainty that
quantifies the distributional uncertainty of a statistical estimand with
respect to Kullback-Liebler divergence, that is, the sensitivity of the
parameter under general distributional perturbations within a Kullback-Liebler
divergence ball. If the signal-to-noise ratio is small, distributional
uncertainty is a monotonous transformation of the signal-to-noise ratio. In
general, however, it is a different concept and corresponds to a different
research question. Further, we propose measures to estimate the stability of
parameters with respect to directional or variable-specific shifts. We also
demonstrate how the measure of distributional uncertainty can be used to
prioritize data collection for better estimation of statistical parameters
under shifted distribution. We evaluate the performance of the proposed measure
in simulations and real data and show that it can elucidate the distributional
(in-)stability of an estimator with respect to certain shifts and give more
accurate estimates of parameters under shifted distribution only requiring to
collect limited information from the shifted distribution.
| stat |
Valid Two-Sample Graph Testing via Optimal Transport Procrustes and
Multiscale Graph Correlation: Applications in Connectomics | Testing whether two graphs come from the same distribution is of interest in
many real world scenarios, including brain network analysis. Under the random
dot product graph model, the nonparametric hypothesis testing framework
consists of embedding the graphs using the adjacency spectral embedding (ASE),
followed by aligning the embeddings using the median flip heuristic, and
finally applying the nonparametric maximum mean discrepancy (MMD) test to
obtain a p-value. Using synthetic data generated from Drosophila brain
networks, we show that the median flip heuristic results in an invalid test,
and demonstrate that optimal transport Procrustes (OTP) for alignment resolves
the invalidity. We further demonstrate that substituting the MMD test with
multiscale graph correlation (MGC) test leads to a more powerful test both in
synthetic and in simulated data. Lastly, we apply this valid and more powerful
test to the right and left hemispheres of the larval Drosophila mushroom body
brain networks, and conclude that there is not enough evidence to reject the
null hypothesis that the two hemispheres are equally distributed.
| stat |
A comparison of Bayesian accelerated failure time models with spatially
varying coefficients | The accelerated failure time (AFT) model is a commonly used tool in analyzing
survival data. In public health studies, data is often collected from medical
service providers in different locations. Survival rates from different
locations often present geographically varying patterns. In this paper, we
focus on the accelerated failure time model with spatially varying
coefficients. We compare three types of the priors for spatially varying
coefficients. A model selection criterion, logarithm of the pseudo-marginal
likelihood (LPML), is developed to assess the fit of AFT model with different
priors. Extensive simulation studies are carried out to examine the empirical
performance of the proposed methods. Finally, we apply our model to SEER data
on prostate cancer in Louisiana and demonstrate the existence of spatially
varying effects on survival rates from prostate cancer data.
| stat |
Scalable Gaussian Process Computations Using Hierarchical Matrices | We present a kernel-independent method that applies hierarchical matrices to
the problem of maximum likelihood estimation for Gaussian processes. The
proposed approximation provides natural and scalable stochastic estimators for
its gradient and Hessian, as well as the expected Fisher information matrix,
that are computable in quasilinear $O(n \log^2 n)$ complexity for a large range
of models. To accomplish this, we (i) choose a specific hierarchical
approximation for covariance matrices that enables the computation of their
exact derivatives and (ii) use a stabilized form of the Hutchinson stochastic
trace estimator. Since both the observed and expected information matrices can
be computed in quasilinear complexity, covariance matrices for MLEs can also be
estimated efficiently. After discussing the associated mathematics, we
demonstrate the scalability of the method, discuss details of its
implementation, and validate that the resulting MLEs and confidence intervals
based on the inverse Fisher information matrix faithfully approach those
obtained by the exact likelihood.
| stat |
Wasserstein Regression | The analysis of samples of random objects that do not lie in a vector space
has found increasing attention in statistics in recent years. An important
class of such object data is univariate probability measures defined on the
real line. Adopting the Wasserstein metric, we develop a class of regression
models for such data, where random distributions serve as predictors and the
responses are either also distributions or scalars. To define this regression
model, we utilize the geometry of tangent bundles of the metric space of random
measures with the Wasserstein metric. The proposed distribution-to-distribution
regression model provides an extension of multivariate linear regression for
Euclidean data and function-to-function regression for Hilbert space valued
data in functional data analysis. In simulations, it performs better than an
alternative approach where one first transforms the distributions to functions
in a Hilbert space and then applies traditional functional regression. We
derive asymptotic rates of convergence for the estimator of the regression
coefficient function and for predicted distributions and also study an
extension to autoregressive models for distribution-valued time series. The
proposed methods are illustrated with data on human mortality and distributions
of house prices.
| stat |
Average Treatment Effect Estimation in Observational Studies with
Functional Covariates | Functional data analysis is an important area in modern statistics and has
been successfully applied in many fields. Although many scientific studies aim
to find causations, a predominant majority of functional data analysis
approaches can only reveal correlations. In this paper, average treatment
effect estimation is studied for observational data with functional covariates.
This paper generalizes various state-of-art propensity score estimation methods
for multivariate data to functional data. The resulting average treatment
effect estimators via propensity score weighting are numerically evaluated by a
simulation study and applied to a real-world dataset to study the causal effect
of duloxitine on the pain relief of chronic knee osteoarthritis patients.
| stat |
Optimal Combination of Linear and Spectral Estimators for Generalized
Linear Models | We study the problem of recovering an unknown signal $\boldsymbol x$ given
measurements obtained from a generalized linear model with a Gaussian sensing
matrix. Two popular solutions are based on a linear estimator $\hat{\boldsymbol
x}^{\rm L}$ and a spectral estimator $\hat{\boldsymbol x}^{\rm s}$. The former
is a data-dependent linear combination of the columns of the measurement
matrix, and its analysis is quite simple. The latter is the principal
eigenvector of a data-dependent matrix, and a recent line of work has studied
its performance. In this paper, we show how to optimally combine
$\hat{\boldsymbol x}^{\rm L}$ and $\hat{\boldsymbol x}^{\rm s}$. At the heart
of our analysis is the exact characterization of the joint empirical
distribution of $(\boldsymbol x, \hat{\boldsymbol x}^{\rm L}, \hat{\boldsymbol
x}^{\rm s})$ in the high-dimensional limit. This allows us to compute the
Bayes-optimal combination of $\hat{\boldsymbol x}^{\rm L}$ and
$\hat{\boldsymbol x}^{\rm s}$, given the limiting distribution of the signal
$\boldsymbol x$. When the distribution of the signal is Gaussian, then the
Bayes-optimal combination has the form $\theta\hat{\boldsymbol x}^{\rm
L}+\hat{\boldsymbol x}^{\rm s}$ and we derive the optimal combination
coefficient. In order to establish the limiting distribution of $(\boldsymbol
x, \hat{\boldsymbol x}^{\rm L}, \hat{\boldsymbol x}^{\rm s})$, we design and
analyze an Approximate Message Passing (AMP) algorithm whose iterates give
$\hat{\boldsymbol x}^{\rm L}$ and approach $\hat{\boldsymbol x}^{\rm s}$.
Numerical simulations demonstrate the improvement of the proposed combination
with respect to the two methods considered separately.
| stat |
Convex Programming for Estimation in Nonlinear Recurrent Models | We propose a formulation for nonlinear recurrent models that includes simple
parametric models of recurrent neural networks as a special case. The proposed
formulation leads to a natural estimator in the form of a convex program. We
provide a sample complexity for this estimator in the case of stable dynamics,
where the nonlinear recursion has a certain contraction property, and under
certain regularity conditions on the input distribution. We evaluate the
performance of the estimator by simulation on synthetic data. These numerical
experiments also suggest the extent at which the imposed theoretical
assumptions may be relaxed.
| stat |
Assessing the Validity of a a priori Patient-Trial Generalizability
Score using Real-world Data from a Large Clinical Data Research Network: A
Colorectal Cancer Clinical Trial Case Study | Existing trials had not taken enough consideration of their population
representativeness, which can lower the effectiveness when the treatment is
applied in real-world clinical practice. We analyzed the eligibility criteria
of Bevacizumab colorectal cancer treatment trials, assessed their a priori
generalizability, and examined how it affects patient outcomes when applied in
real-world clinical settings. To do so, we extracted patient-level data from a
large collection of electronic health records (EHRs) from the OneFlorida
consortium. We built a zero-inflated negative binomial model using a composite
patient-trial generalizability (cPTG) score to predict patients clinical
outcomes (i.e., number of serious adverse events, (SAEs)). Our study results
provide a body of evidence that 1) the cPTG scores can predict patient
outcomes; and 2) patients who are more similar to the study population in the
trials that were used to develop the treatment will have a significantly lower
possibility to experience serious adverse events.
| stat |
Monte Carlo simulation studies on Python using the sstudy package with
SQL databases as storage | Performance assessment is a key issue in the process of proposing new machine
learning/statistical estimators. A possible method to complete such task is by
using simulation studies, which can be defined as the procedure of estimating
and comparing properties (such as predictive power) of estimators (and other
statistics) by averaging over many replications given a true distribution;
i.e.: generating a dataset, fitting the estimator, calculating and storing the
predictive power, and then repeating the procedure many times and finally
averaging over the stored predictive powers. Given that, in this paper, we
present sstudy: a Python package designed to simplify the preparation of
simulation studies using SQL database engines as the storage system; more
specifically, we present its basic features, usage examples and references to
the its documentation. We also present a short statistical description of the
simulation study procedure with a simplified explanation of what is being
estimated by it, as well as some examples of applications.
| stat |
Prediction Intervals for Synthetic Control Methods | Uncertainty quantification is a fundamental problem in the analysis and
interpretation of synthetic control (SC) methods. We develop conditional
prediction intervals in the SC framework, and provide conditions under which
these intervals offer finite-sample probability guarantees. Our method allows
for covariate adjustment and non-stationary data, among other practically
relevant features. The construction begins by noting that the statistical
uncertainty of the SC prediction is governed by two distinct sources of
randomness: one coming from the construction of the (likely misspecified) SC
weights in the pre-treatment period, and the other coming from the unobservable
stochastic error in the post-treatment period when the treatment effect is
analyzed. Accordingly, our proposed prediction intervals are constructed taking
into account both sources of randomness. For implementation, we propose a
simulation-based approach along with finite-sample-based probability bound
arguments, naturally leading to principled sensitivity analysis methods. We
illustrate the numerical performance of our methods using empirical
applications and a small simulation study.
| stat |
Estimating the travel time and the most likely path from Lagrangian
drifters | We provide a novel methodology for computing the most likely path taken by
drifters between arbitrary fixed locations in the ocean. We also provide an
estimate of the travel time associated with this path. Lagrangian pathways and
travel times are of practical value not just in understanding surface
velocities, but also in modelling the transport of ocean-borne species such as
planktonic organisms, and floating debris such as plastics. In particular, the
estimated travel time can be used to compute an estimated Lagrangian distance,
which is often more informative than Euclidean distance in understanding
connectivity between locations. Our methodology is purely data-driven, and
requires no simulations of drifter trajectories, in contrast to existing
approaches. Our method scales globally and can simultaneously handle multiple
locations in the ocean. Furthermore, we provide estimates of the error and
uncertainty associated with both the most likely path and the associated travel
time.
| stat |
Posterior Averaging Information Criterion | We propose a new model selection method, the posterior averaging information
criterion, for Bayesian model assessment from a predictive perspective. The
theoretical foundation is built on the Kullback-Leibler divergence to quantify
the similarity between the proposed candidate model and the underlying true
model. From a Bayesian perspective, our method evaluates the candidate models
over the entire posterior distribution in terms of predicting a future
independent observation. Without assuming that the true distribution is
contained in the candidate models, the new criterion is developed by correcting
the asymptotic bias of the posterior mean of the log-likelihood against its
expected log-likelihood. It can be generally applied even for Bayesian models
with degenerate non-informative prior. The simulation in both normal and
binomial settings demonstrates decent small sample performance.
| stat |
A bi-partite generative model framework for analyzing and simulating
large scale multiple discrete-continuous travel behaviour data | The emergence of data-driven demand analysis has led to the increased use of
generative modelling to learn the probabilistic dependencies between random
variables. Although their apparent use has mostly been limited to image
recognition and classification in recent years, generative machine learning
algorithms can be a powerful tool for travel behaviour research by replicating
travel behaviour by the underlying properties of data structures. In this
paper, we examine the use of generative machine learning approach for analyzing
multiple discrete-continuous (MDC) travel behaviour data. We provide a
plausible perspective of how we can exploit the use of machine learning
techniques to interpret the underlying heterogeneities in the data. We show
that generative models are conceptually similar to the choice selection
behaviour process through information entropy and variational Bayesian
inference. Without loss of generality, we consider a restricted Boltzmann
machine (RBM) based algorithm with multiple discrete-continuous layers,
formulated as a variational Bayesian inference optimization problem. We
systematically describe the proposed machine learning algorithm and develop a
process of analyzing travel behaviour data from a generative learning
perspective. We show parameter stability from model analysis and simulation
tests on an open dataset with multiple discrete-continuous dimensions from a
data size of 293,330 observations. For interpretability, we derive the
conditional probabilities, elasticities and perform statistical analysis on the
latent variables. We show that our model can generate statistically similar
data distributions for travel forecasting and prediction and performs better
than purely discriminative methods in validation. Our results indicate that
latent constructs in generative models can accurately represent the joint
distribution consistently on MDC data.
| stat |
On the Convergence of FedAvg on Non-IID Data | Federated learning enables a large amount of edge computing devices to
jointly learn a model without data sharing. As a leading algorithm in this
setting, Federated Averaging (\texttt{FedAvg}) runs Stochastic Gradient Descent
(SGD) in parallel on a small subset of the total devices and averages the
sequences only once in a while. Despite its simplicity, it lacks theoretical
guarantees under realistic settings. In this paper, we analyze the convergence
of \texttt{FedAvg} on non-iid data and establish a convergence rate of
$\mathcal{O}(\frac{1}{T})$ for strongly convex and smooth problems, where $T$
is the number of SGDs. Importantly, our bound demonstrates a trade-off between
communication-efficiency and convergence rate. As user devices may be
disconnected from the server, we relax the assumption of full device
participation to partial device participation and study different averaging
schemes; low device participation rate can be achieved without severely slowing
down the learning. Our results indicate that heterogeneity of data slows down
the convergence, which matches empirical observations. Furthermore, we provide
a necessary condition for \texttt{FedAvg} on non-iid data: the learning rate
$\eta$ must decay, even if full-gradient is used; otherwise, the solution will
be $\Omega (\eta)$ away from the optimal.
| stat |
Fully reconciled GDP forecasts from Income and Expenditure sides | We propose a complete reconciliation procedure, resulting in a 'one number
forecast' of the GDP figure, coherent with both Income and Expenditure sides'
forecasted series, and evaluate its performance on the Australian quarterly GDP
series, as compared to the original proposal by Athanasopoulos et al. (2019).
| stat |
Understanding the Behaviour of the Empirical Cross-Entropy Beyond the
Training Distribution | Machine learning theory has mostly focused on generalization to samples from
the same distribution as the training data. Whereas a better understanding of
generalization beyond the training distribution where the observed distribution
changes is also fundamentally important to achieve a more powerful form of
generalization. In this paper, we attempt to study through the lens of
information measures how a particular architecture behaves when the true
probability law of the samples is potentially different at training and testing
times. Our main result is that the testing gap between the empirical
cross-entropy and its statistical expectation (measured with respect to the
testing probability law) can be bounded with high probability by the mutual
information between the input testing samples and the corresponding
representations, generated by the encoder obtained at training time. These
results of theoretical nature are supported by numerical simulations showing
that the mentioned mutual information is representative of the testing gap,
capturing qualitatively the dynamic in terms of the hyperparameters of the
network.
| stat |
Implementation of Frequency-Severity Association in BMS Ratemaking | A Bonus-Malus System (BMS) in insurance is a premium adjustment mechanism
widely used in a posteriori ratemaking process to set the premium for the next
contract period based on a policyholder's claim history. The current practice
in BMS implementation relies on the assumption of independence between claim
frequency and severity, despite the fact that a series of recent studies report
evidence of a significant frequency-severity relationship, particularly in
automobile insurance. To address this discrepancy, we propose a copula-based
correlated random effects model to accommodate the dependence between claim
frequency and severity, and further illustrate how to incorporate such
dependence into the current BMS. We derive analytical solutions to the optimal
relativities under the proposed framework and provide numerical experiments
based on real data analysis to assess the effect of frequency-severity
dependence in BMS ratemaking.
| stat |
Poisson Network Autoregression | We consider network autoregressive models for count data with a non-random
time-varying neighborhood structure. The main methodological contribution is
the development of conditions that guarantee stability and valid statistical
inference. We consider both cases of fixed and increasing network dimension and
we show that quasi-likelihood inference provides consistent and asymptotically
normally distributed estimators. The work is complemented by simulation results
and a data example.
| stat |
Future Proofing a Building Design Using History Matching Inspired
Level-Set Techniques | History Matching is a technique used to calibrate complex computer models,
that is, finding the input settings which lead to the simulated output matching
up with real world observations. Key to this technique is the construction of
emulators, which provide fast probabilistic predictions of future simulations.
In this work, we adapt the History Matching framework to tackle the problem of
level set estimation, that is, finding input settings where the output is below
(or above) some threshold. The developed methodology is heavily motivated by a
specific case study: how can one design a building that will be sufficiently
protected against overheating and sufficiently energy efficient, whilst
considering the expected increases in temperature due to climate change? We
successfully manage to address this - greatly reducing a large initial set of
candidate building designs down to a small set of acceptable potential
buildings.
| stat |
Simulation Framework for Realistic Large-scale Individual-level Health
Data Generation | We propose a general framework for realistic data generation and simulation
of complex systems in the health domain. The main use cases of the framework
are predicting the development of risk factors and disease occurrence,
evaluating the impact of interventions and policy decisions, and statistical
method development. We present the fundamentals of the framework using rigorous
mathematical definitions. The framework supports calibration to a real
population as well as various manipulations and data collection processes. The
freely available open-source implementation in R embraces efficient data
structures, parallel computing and fast random number generation which ensure
reproducibility and scalability. With the framework it is possible to run
daily-level simulations for populations of millions individuals for decades of
simulated time. An example on the occurrence of stroke, type 2 diabetes and
mortality illustrates the usage of the framework in the Finnish context. In the
example, we demonstrate the data-collection functionality by studying the
impact of non-participation on the estimated risk models.
| stat |
Bounds for the weight of external data in shrinkage estimation | Shrinkage estimation in a meta-analysis framework may be used to facilitate
dynamical borrowing of information. This framework might be used to analyze a
new study in the light of previous data, which might differ in their design
(e.g., a randomized controlled trial (RCT) and a clinical registry). We show
how the common study weights arise in effect and shrinkage estimation, and how
these may be generalized to the case of Bayesian meta-analysis. Next we develop
simple ways to compute bounds on the weights, so that the contribution of the
external evidence may be assessed a priori. These considerations are
illustrated and discussed using numerical examples, including applications in
the treatment of Creutzfeldt-Jakob disease and in fetal monitoring to prevent
the occurrence of metabolic acidosis. The target study's contribution to the
resulting estimate is shown to be bounded below. Therefore, concerns of
evidence being easily overwhelmed by external data are largely unwarranted.
| 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.