title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
mgm: Estimating Time-Varying Mixed Graphical Models in High-Dimensional
Data | We present the R-package mgm for the estimation of k-order Mixed Graphical
Models (MGMs) and mixed Vector Autoregressive (mVAR) models in high-dimensional
data. These are a useful extensions of graphical models for only one variable
type, since data sets consisting of mixed types of variables (continuous,
count, categorical) are ubiquitous. In addition, we allow to relax the
stationarity assumption of both models by introducing time-varying versions
MGMs and mVAR models based on a kernel weighting approach. Time-varying models
offer a rich description of temporally evolving systems and allow to identify
external influences on the model structure such as the impact of interventions.
We provide the background of all implemented methods and provide fully
reproducible examples that illustrate how to use the package.
| stat |
An automatic procedure to determine groups of nonparametric regression
curves | In many situations it could be interesting to ascertain whether nonparametric
regression curves can be grouped, especially when confronted with a
considerable number of curves. The proposed testing procedure allows to
determine groups with an automatic selection of their number. A simulation
study is presented in order to investigate the finite sample properties of the
proposed methods when compared to existing alternative procedures. Finally, the
applicability of the procedure to study the geometry of a tunnel by analysing a
set of cross-sections is demonstrated. The results obtained show the existence
of some heterogeneity in the tunnel geometry.
| stat |
Variational Representations and Neural Network Estimation of R\'enyi
Divergences | We derive a new variational formula for the R{\'e}nyi family of divergences,
$R_\alpha(Q\|P)$, between probability measures $Q$ and $P$. Our result
generalizes the classical Donsker-Varadhan variational formula for the
Kullback-Leibler divergence. We further show that this R{\'e}nyi variational
formula holds over a range of function spaces; this leads to a formula for the
optimizer under very weak assumptions and is also key in our development of a
consistency theory for R{\'e}nyi divergence estimators. By applying this theory
to neural-network estimators, we show that if a neural network family satisfies
one of several strengthened versions of the universal approximation property
then the corresponding R{\'e}nyi divergence estimator is consistent. In
contrast to density-estimator based methods, our estimators involve only
expectations under $Q$ and $P$ and hence are more effective in high dimensional
systems. We illustrate this via several numerical examples of neural network
estimation in systems of up to 5000 dimensions.
| stat |
How response designs and class proportions affect the accuracy of
validation data | Reference data collected to validate land cover maps are generally considered
free of errors. In practice, however, they contain errors despite all efforts
to minimise them. These errors then propagate up to the accuracy assessment
stage and impact the validation results. For photo-interpreted reference data,
the three most widely studied sources of error are systematic incorrect
labelling, vigilance drops, and demographic factors. How internal estimation
errors, i.e., errors intrinsic to the response design, affect the accuracy of
reference data is far less understood. We analysed the impact of estimation
errors for two types of legends as well as for point-based and partition-based
response designs with a range of sub-sample sizes. We showed that the accuracy
of response designs depends on the class proportions within the sampling units,
with complex landscapes being more prone to errors. As a result, response
designs where the number of sub-samples are fixed are inefficient, and the
labels of reference data sets have inconsistent confidence levels. To control
estimation errors, to guarantee high accuracy standards of validation data, and
to minimise data collection efforts, we proposed to rely on confidence
intervals of the photo-interpreted data to define how many sub-samples should
be labelled. In practice, sub-samples are iteratively selected and labelled
until the estimated class proportions reach the desired level of confidence. As
a result, less effort is spent on labelling obvious cases and the spared effort
can be allocated to more complex cases. This approach could reduce the
labelling effort by 50% to 75% in our study site, with greater gains in
homogeneous landscapes. We contend that adopting this optimisation approach
will not only increase the efficiency of reference data collection but will
also help deliver reliable accuracy estimates to the user community.
| stat |
Adaptive quadrature schemes for Bayesian inference via active learning | Numerical integration and emulation are fundamental topics across scientific
fields. We propose novel adaptive quadrature schemes based on an active
learning procedure. We consider an interpolative approach for building a
surrogate posterior density, combining it with Monte Carlo sampling methods and
other quadrature rules. The nodes of the quadrature are sequentially chosen by
maximizing a suitable acquisition function, which takes into account the
current approximation of the posterior and the positions of the nodes. This
maximization does not require additional evaluations of the true posterior. We
introduce two specific schemes based on Gaussian and Nearest Neighbors (NN)
bases. For the Gaussian case, we also provide a novel procedure for fitting the
bandwidth parameter, in order to build a suitable emulator of a density
function. With both techniques, we always obtain a positive estimation of the
marginal likelihood (a.k.a., Bayesian evidence). An equivalent importance
sampling interpretation is also described, which allows the design of extended
schemes. Several theoretical results are provided and discussed. Numerical
results show the advantage of the proposed approach, including a challenging
inference problem in an astronomic dynamical model, with the goal of revealing
the number of planets orbiting a star.
| stat |
Modelling age-related changes in executive functions of soccer players | The widespread popularity of soccer across the globe has turned it into a
multi-billion dollar industry. As a result, most professional clubs actively
engage in talent identification and development programmes. Contemporary
research has generally supported the use of executive functions - a class of
neuropsychological processes responsible for cognitive behaviours - in
predicting a soccer player's future success. However, studies on the
developmental evolution of executive functions have yielded differing results
in their structural form (such as inverted U-shapes, or otherwise). This
article presents the first analysis of changes in the domain-generic and
domain-specific executive functions based on longitudinal data measured on
elite German soccer players. Results obtained from a latent variable model show
that these executive functions experience noticeable growth from late childhood
until pre-adolescence, but remain fairly stable in later growth stages. As a
consequence, our results suggest that the role of executive functions in
facilitating talent identification may have been overly emphasised.
| stat |
Bayesian shared-parameter models for analysing sardine fishing in the
Mediterranean Sea | European sardine is experiencing an overfishing around the world. The
dynamics of the industrial and artisanal fishing in the Mediterranean Sea from
1970 to 2014 by country was assessed by means of Bayesian joint longitudinal
modelling that uses the random effects to generate an association structure
between both longitudinal measures. Model selection was based on Bayes factors
approximated through the harmonic mean.
| stat |
When Does Preconditioning Help or Hurt Generalization? | While second order optimizers such as natural gradient descent (NGD) often
speed up optimization, their effect on generalization has been called into
question. This work presents a more nuanced view on how the \textit{implicit
bias} of first- and second-order methods affects the comparison of
generalization properties. We provide an exact asymptotic bias-variance
decomposition of the generalization error of overparameterized ridgeless
regression under a general class of preconditioner $\boldsymbol{P}$, and
consider the inverse population Fisher information matrix (used in NGD) as a
particular example. We determine the optimal $\boldsymbol{P}$ for both the bias
and variance, and find that the relative generalization performance of
different optimizers depends on the label noise and the "shape" of the signal
(true parameters): when the labels are noisy, the model is misspecified, or the
signal is misaligned with the features, NGD can achieve lower risk; conversely,
GD generalizes better than NGD under clean labels, a well-specified model, or
aligned signal. Based on this analysis, we discuss several approaches to manage
the bias-variance tradeoff, and the potential benefit of interpolating between
GD and NGD. We then extend our analysis to regression in the reproducing kernel
Hilbert space and demonstrate that preconditioned GD can decrease the
population risk faster than GD. Lastly, we empirically compare the
generalization error of first- and second-order optimizers in neural network
experiments, and observe robust trends matching our theoretical analysis.
| stat |
Identifiability of Generalized Hypergeometric Distribution (GHD)
Directed Acyclic Graphical Models | We introduce a new class of identifiable DAG models where the conditional
distribution of each node given its parents belongs to a family of generalized
hypergeometric distributions (GHD). A family of generalized hypergeometric
distributions includes a lot of discrete distributions such as the binomial,
Beta-binomial, negative binomial, Poisson, hyper-Poisson, and many more. We
prove that if the data drawn from the new class of DAG models, one can fully
identify the graph structure. We further present a reliable and polynomial-time
algorithm that recovers the graph from finitely many data. We show through
theoretical results and numerical experiments that our algorithm is
statistically consistent in high-dimensional settings (p>n) if the indegree of
the graph is bounded, and out-performs state-of-the-art DAG learning
algorithms.
| stat |
MM Algorithms for Distance Covariance based Sufficient Dimension
Reduction and Sufficient Variable Selection | Sufficient dimension reduction (SDR) using distance covariance (DCOV) was
recently proposed as an approach to dimension-reduction problems. Compared with
other SDR methods, it is model-free without estimating link function and does
not require any particular distributions on predictors (see Sheng and Yin,
2013, 2016). However, the DCOV-based SDR method involves optimizing a nonsmooth
and nonconvex objective function over the Stiefel manifold. To tackle the
numerical challenge, we novelly formulate the original objective function
equivalently into a DC (Difference of Convex functions) program and construct
an iterative algorithm based on the majorization-minimization (MM) principle.
At each step of the MM algorithm, we inexactly solve the quadratic subproblem
on the Stiefel manifold by taking one iteration of Riemannian Newton's method.
The algorithm can also be readily extended to sufficient variable selection
(SVS) using distance covariance. We establish the convergence property of the
proposed algorithm under some regularity conditions. Simulation studies show
our algorithm drastically improves the computation efficiency and is robust
across various settings compared with the existing method. Supplemental
materials for this article are available.
| stat |
Unsupervised Contextual Anomaly Detection using Joint Deep Variational
Generative Models | A method for unsupervised contextual anomaly detection is proposed using a
cross-linked pair of Variational Auto-Encoders for assigning a normality score
to an observation. The method enables a distinct separation of contextual from
behavioral attributes and is robust to the presence of anomalous or novel
contextual attributes. The method can be trained with data sets that contain
anomalies without any special pre-processing.
| stat |
Dynamic prediction of time to event with survival curves | With the ever-growing complexity of primary health care system, proactive
patient failure management is an effective way to enhancing the availability of
health care resource. One key enabler is the dynamic prediction of
time-to-event outcomes. Conventional explanatory statistical approach lacks the
capability of making precise individual level prediction, while the data
adaptive binary predictors does not provide nominal survival curves for
biologically plausible survival analysis. The purpose of this article is to
elucidate that the knowledge of explanatory survival analysis can significantly
enhance the current black-box data adaptive prediction models. We apply our
recently developed counterfactual dynamic survival model (CDSM) to static and
longitudinal observational data and testify that the inflection point of its
estimated individual survival curves provides reliable prediction of the
patient failure time.
| stat |
Queueing Subject To Action-Dependent Server Performance: Utilization
Rate Reduction | We consider a discrete-time system comprising a first-come-first-served
queue, a non-preemptive server, and a stationary non-work-conserving scheduler.
New tasks enter the queue according to a Bernoulli process with a pre-specified
arrival rate. At each instant, the server is either busy working on a task or
is available. When the server is available, the scheduler either assigns a new
task to the server or allows it to remain available (to rest). In addition to
the aforementioned availability state, we assume that the server has an
integer-valued activity state. The activity state is non-decreasing during work
periods, and is non-increasing otherwise. In a typical application of our
framework, the server performance (understood as task completion probability)
worsens as the activity state increases. In this article, we build on and
transcend recent stabilizability results obtained for the same framework.
Specifically, we establish methods to design scheduling policies that not only
stabilize the queue but also reduce the utilization rate - understood as the
infinite-horizon time-averaged portion of time the server is working. This
article has a main theorem leading to two key results: (i) We put forth a
tractable method to determine, using a finite-dimensional linear program (LP),
the infimum of all utilization rates that can be achieved by scheduling
policies that are stabilizing, for a given arrival rate. (ii) We propose a
design method, also based on finite-dimensional LPs, to obtain stabilizing
scheduling policies that can attain a utilization rate arbitrarily close to the
aforementioned infimum. We also establish structural and distributional
convergence properties, which are used throughout the article, and are
significant in their own right.
| stat |
Observational daily and regional photovoltaic solar energy production
for the Netherlands | This paper presents a classical estimation problem for calculating the energy
generated by photovoltaic solar energy systems in the Netherlands, on a daily,
annual and regional basis. We identify two data sources to construct our
methodology: pvoutput, an online portal with solar energy yield measurements,
and modelled irradiance data, from the Royal Netherlands Meteorological
Institute. Combining these, we obtain probability functions of observing energy
yields, given the irradiance, which we then apply to our PV systems database,
allowing us to calculate daily and annual solar energy yields. We examine the
variation in our daily and annual estimates as a result of taking different
subsets of pvoutput systems with certain specifications such as orientation,
tilt and inverter to PV capacity ratio. Hence we obtain specific annual energy
yields in the range of 877-946 kWh/kWp and 838-899 kWh/kWp for 2016 and 2017
respectively. The current method used at Statistics Netherlands assumes this to
be 875 kWh/kWp, meaning the yields were underestimated and overestimated for
2016 and 2017 respectively. Finally, we translate our national estimates into
solar energy yields per municipality. This research demonstrates that an
irradiance based measure of solar energy generation is necessary to obtain more
accurate energy yields on both a national and regional level.
| stat |
More efficient approximation of smoothing splines via space-filling
basis selection | We consider the problem of approximating smoothing spline estimators in a
nonparametric regression model. When applied to a sample of size $n$, the
smoothing spline estimator can be expressed as a linear combination of $n$
basis functions, requiring $O(n^3)$ computational time when the number of
predictors $d\geq 2$. Such a sizable computational cost hinders the broad
applicability of smoothing splines. In practice, the full sample smoothing
spline estimator can be approximated by an estimator based on $q$
randomly-selected basis functions, resulting in a computational cost of
$O(nq^2)$. It is known that these two estimators converge at the identical rate
when $q$ is of the order $O\{n^{2/(pr+1)}\}$, where $p\in [1,2]$ depends on the
true function $\eta$, and $r > 1$ depends on the type of spline. Such $q$ is
called the essential number of basis functions. In this article, we develop a
more efficient basis selection method. By selecting the ones corresponding to
roughly equal-spaced observations, the proposed method chooses a set of basis
functions with a large diversity. The asymptotic analysis shows our proposed
smoothing spline estimator can decrease $q$ to roughly $O\{n^{1/(pr+1)}\}$,
when $d\leq pr+1$. Applications on synthetic and real-world datasets show the
proposed method leads to a smaller prediction error compared with other basis
selection methods.
| stat |
On spectral algorithms for community detection in stochastic blockmodel
graphs with vertex covariates | In network inference applications, it is often desirable to detect community
structure, namely to cluster vertices into groups, or blocks, according to some
measure of similarity. Beyond mere adjacency matrices, many real networks also
involve vertex covariates that carry key information about underlying block
structure in graphs. To assess the effects of such covariates on block
recovery, we present a comparative analysis of two model-based spectral
algorithms for clustering vertices in stochastic blockmodel graphs with vertex
covariates. The first algorithm uses only the adjacency matrix, and directly
estimates the induced block assignments. The second algorithm incorporates both
the adjacency matrix and the vertex covariates into the estimation of block
assignments, and moreover quantifies the explicit impact of the vertex
covariates on the resulting estimate of the block assignments. We employ
Chernoff information to analytically compare the algorithms' performance and
derive the Chernoff ratio for certain models of interest. Analytic results and
simulations suggest that the second algorithm is often preferred: we can often
better estimate the induced block assignments by first estimating the effect of
vertex covariates. In addition, real data examples on diffusion MRI connectome
datasets and social network datasets also indicate that the second algorithm
has the advantages of revealing underlying block structure and taking observed
vertex heterogeneity into account in real applications. Our findings emphasize
the importance of distinguishing between observed and unobserved factors that
can affect block structure in graphs.
| stat |
A spectral adjustment for spatial confounding | Adjusting for an unmeasured confounder is generally an intractable problem,
but in the spatial setting it may be possible under certain conditions. In this
paper, we derive necessary conditions on the coherence between the treatment
variable of interest and the unmeasured confounder that ensure the causal
effect of the treatment is estimable. We specify our model and assumptions in
the spectral domain to allow for different degrees of confounding at different
spatial resolutions. The key assumption that ensures identifiability is that
confounding present at global scales dissipates at local scales. We show that
this assumption in the spectral domain is equivalent to adjusting for
global-scale confounding in the spatial domain by adding a spatially smoothed
version of the treatment variable to the mean of the response variable. Within
this general framework, we propose a sequence of confounder adjustment methods
that range from parametric adjustments based on the Matern coherence function
to more robust semi-parametric methods that use smoothing splines. These ideas
are applied to areal and geostatistical data for both simulated and real
datasets
| stat |
Amortized Bayesian Inference for Models of Cognition | As models of cognition grow in complexity and number of parameters, Bayesian
inference with standard methods can become intractable, especially when the
data-generating model is of unknown analytic form. Recent advances in
simulation-based inference using specialized neural network architectures
circumvent many previous problems of approximate Bayesian computation.
Moreover, due to the properties of these special neural network estimators, the
effort of training the networks via simulations amortizes over subsequent
evaluations which can re-use the same network for multiple datasets and across
multiple researchers. However, these methods have been largely underutilized in
cognitive science and psychology so far, even though they are well suited for
tackling a wide variety of modeling problems. With this work, we provide a
general introduction to amortized Bayesian parameter estimation and model
comparison and demonstrate the applicability of the proposed methods on a
well-known class of intractable response-time models.
| stat |
Sampling from Networks: Respondent-Driven Sampling | Respondent-Driven Sampling (RDS) is a variant of link-tracing, a sampling
technique for surveying hard-to-reach communities that takes advantage of
community members' social networks to reach potential participants. As a
network-based sampling method, RDS is faced with the fundamental problem of
sampling from population networks where features such as homophily (the
tendency for individuals with similar traits to share social ties) and
differential activity (the ratio of the average number of connections by
attribute) are sensitive to the choice of a sampling method. Though not clearly
described in the RDS literature, many simple methods exist to generate
simulated RDS data, with specific levels of network features, where the focus
is on estimating simple estimands. However, the accuracy of these methods in
their abilities to consistently recover those targeted network features remains
unclear. This is also motivated by recent findings that some population network
parameters (e.g.~homophily) cannot be consistently estimated from the RDS data
alone \citep{Crawford17}.
In this paper, we conduct a simulation study to assess the accuracy of
existing RDS simulation methods, in terms of their abilities to generate RDS
samples with the desired levels of two network parameters: homophily and
differential activity. The results show that (1) homophily cannot be
consistently estimated from simulated RDS samples and (2) differential activity
estimates are more precise when groups, defined by traits, are equally active
and equally represented in the population. We use this approach to mimic
features of the Engage Study, an RDS sample of gay, bisexual and other men who
have sex with men in Montreal.
| stat |
An Exploratory Analysis of Biased Learners in Soft-Sensing Frames | Data driven soft sensor design has recently gained immense popularity, due to
advances in sensory devices, and a growing interest in data mining. While
partial least squares (PLS) is traditionally used in the process literature for
designing soft sensors, the statistical literature has focused on sparse
learners, such as Lasso and relevance vector machine (RVM), to solve the high
dimensional data problem. In the current study, predictive performances of
three regression techniques, PLS, Lasso and RVM were assessed and compared
under various offline and online soft sensing scenarios applied on datasets
from five real industrial plants, and a simulated process. In offline learning,
predictions of RVM and Lasso were found to be superior to those of PLS when a
large number of time-lagged predictors were used. Online prediction results
gave a slightly more complicated picture. It was found that the minimum
prediction error achieved by PLS under moving window (MW), or just-in-time
learning scheme was decreased up to ~5-10% using Lasso, or RVM. However, when a
small MW size was used, or the optimum number of PLS components was as low as
~1, prediction performance of PLS surpassed RVM, which was found to yield
occasional unstable predictions. PLS and Lasso models constructed via online
parameter tuning generally did not yield better predictions compared to those
constructed via offline tuning. We present evidence to suggest that retaining a
large portion of the available process measurement data in the predictor
matrix, instead of preselecting variables, would be more advantageous for
sparse learners in increasing prediction accuracy. As a result, Lasso is
recommended as a better substitute for PLS in soft sensors; while performance
of RVM should be validated before online application.
| stat |
Bayesian Estimation of Attribute Disclosure Risks in Synthetic Data with
the $\texttt{AttributeRiskCalculation}$ R Package | Synthetic data is a promising approach to privacy protection in many
contexts. A Bayesian synthesis model, also known as a synthesizer, simulates
synthetic values of sensitive variables from their posterior predictive
distributions. The resulting synthetic data can then be released in place of
the confidential data. An important evaluation prior to synthetic data release
is its level of privacy protection, which is often in the form of disclosure
risks evaluation. Attribute disclosure, referring to an intruder correctly
inferring the confidential values of synthetic records, is one type of
disclosure that is challenging to be computationally evaluated. In this paper,
we review and discuss in detail some Bayesian estimation approaches to
attribute disclosure risks evaluation, with examples of commonly-used Bayesian
synthesizers. We create the $\texttt{AttributeRiskCalculation}$ R package to
facilitate its implementation, and demonstrate its functionality with examples
of evaluating attribute disclosure risks in synthetic samples of the Consumer
Expenditure Surveys.
| stat |
Exact Inference with Approximate Computation for Differentially Private
Data via Perturbations | Differential privacy protects individuals' confidential information by
subjecting data summaries to probabilistic perturbation mechanisms, carefully
designed to minimize undue sacrifice of statistical efficiency. When properly
accounted for, differentially private data are conducive to exact inference
when approximate computation techniques are employed. This paper shows that
approximate Bayesian computation, a practical suite of methods to simulate from
approximate posterior distributions of complex Bayesian models, produces exact
posterior samples when applied to differentially private perturbation data. An
importance sampling implementation of Monte Carlo expectation-maximization for
likelihood inference is also discussed. The results illustrate a duality
between approximate computation on exact data, and exact computation on
approximate data. A cleverly designed inferential procedure exploits the
alignment between the statistical tradeoff of privacy versus efficiency, and
the computational tradeoff of approximation versus exactness, so that paying
the cost of one gains the benefit of both.
| stat |
Does preregistration improve the credibility of research findings? | Preregistration entails researchers registering their planned research
hypotheses, methods, and analyses in a time-stamped document before they
undertake their data collection and analyses. This document is then made
available with the published research report to allow readers to identify
discrepancies between what the researchers originally planned to do and what
they actually ended up doing. This historical transparency is supposed to
facilitate judgments about the credibility of the research findings. The
present article provides a critical review of 17 of the reasons behind this
argument. The article covers issues such as HARKing, multiple testing,
p-hacking, forking paths, optional stopping, researchers' biases, selective
reporting, test severity, publication bias, and replication rates. It is
concluded that preregistration's historical transparency does not facilitate
judgments about the credibility of research findings when researchers provide
contemporary transparency in the form of (a) clear rationales for current
hypotheses and analytical approaches, (b) public access to research data,
materials, and code, and (c) demonstrations of the robustness of research
conclusions to alternative interpretations and analytical approaches.
| stat |
Multiple Imputation with Massive Data: An Application to the Panel Study
of Income Dynamics | Multiple imputation (MI) is a popular and well-established method for
handling missing data in multivariate data sets, but its practicality for use
in massive and complex data sets has been questioned. One such data set is the
Panel Study of Income Dynamics (PSID), a longstanding and extensive survey of
household income and wealth in the United States. Missing data for this survey
are currently handled using traditional hot deck methods. We use a sequential
regression/ chained-equation approach, using the software IVEware, to multiply
impute cross-sectional wealth data in the 2013 PSID, and compare analyses of
the resulting imputed data with results from the current hot deck approach.
Practical difficulties, such as non-normally distributed variables, skip
patterns, categorical variables with many levels, and multicollinearity, are
described together with our approaches to overcoming them. We evaluate the
imputation quality and validity with internal diagnostics and external
benchmarking data. MI produces improvements over the existing hot deck approach
by helping preserve correlation structures with efficiency gains. We recommend
the practical implementation of MI and expect greater gains when the fraction
of missing information is large.
| stat |
Computing Estimators of Dantzig Selector type via Column and Constraint
Generation | We consider a class of linear-programming based estimators in reconstructing
a sparse signal from linear measurements. Specific formulations of the
reconstruction problem considered here include Dantzig selector, basis pursuit
(for the case in which the measurements contain no errors), and the fused
Dantzig selector (for the case in which the underlying signal is piecewise
constant). In spite of being estimators central to sparse signal processing and
machine learning, solving these linear programming problems for large scale
instances remains a challenging task, thereby limiting their usage in practice.
We show that classic constraint- and column-generation techniques from large
scale linear programming, when used in conjunction with a commercial
implementation of the simplex method, and initialized with the solution from a
closely-related Lasso formulation, yields solutions with high efficiency in
many settings.
| stat |
Hyperparameters and Tuning Strategies for Random Forest | The random forest algorithm (RF) has several hyperparameters that have to be
set by the user, e.g., the number of observations drawn randomly for each tree
and whether they are drawn with or without replacement, the number of variables
drawn randomly for each split, the splitting rule, the minimum number of
samples that a node must contain and the number of trees. In this paper, we
first provide a literature review on the parameters' influence on the
prediction performance and on variable importance measures.
It is well known that in most cases RF works reasonably well with the default
values of the hyperparameters specified in software packages. Nevertheless,
tuning the hyperparameters can improve the performance of RF. In the second
part of this paper, after a brief overview of tuning strategies we demonstrate
the application of one of the most established tuning strategies, model-based
optimization (MBO). To make it easier to use, we provide the tuneRanger R
package that tunes RF with MBO automatically. In a benchmark study on several
datasets, we compare the prediction performance and runtime of tuneRanger with
other tuning implementations in R and RF with default hyperparameters.
| stat |
Continuous Assortment Optimization with Logit Choice Probabilities under
Incomplete Information | We consider assortment optimization over a continuous spectrum of products
represented by the unit interval, where the seller's problem consists of
determining the optimal subset of products to offer to potential customers. To
describe the relation between assortment and customer choice, we propose a
probabilistic choice model that forms the continuous counterpart of the widely
studied discrete multinomial logit model. We consider the seller's problem
under incomplete information, propose a stochastic-approximation type of
policy, and show that its regret -- its performance loss compared to the
optimal policy -- is only logarithmic in the time horizon. We complement this
result by showing a matching lower bound on the regret of any policy, implying
that our policy is asymptotically optimal. We then show that adding a capacity
constraint significantly changes the structure of the problem: we construct a
policy and show that its regret after $T$ time periods is bounded above by a
constant times $T^{2/3}$ (up to a logarithmic term); in addition, we show that
the regret of any policy is bounded from below by a positive constant times
$T^{2/3}$, so that also in the capacitated case we obtain asymptotic
optimality. Numerical illustrations show that our policies outperform or are on
par with alternatives.
| stat |
Symmetry-adapted generation of 3d point sets for the targeted discovery
of molecules | Deep learning has proven to yield fast and accurate predictions of
quantum-chemical properties to accelerate the discovery of novel molecules and
materials. As an exhaustive exploration of the vast chemical space is still
infeasible, we require generative models that guide our search towards systems
with desired properties. While graph-based models have previously been
proposed, they are restricted by a lack of spatial information such that they
are unable to recognize spatial isomerism and non-bonded interactions. Here, we
introduce a generative neural network for 3d point sets that respects the
rotational invariance of the targeted structures. We apply it to the generation
of molecules and demonstrate its ability to approximate the distribution of
equilibrium structures using spatial metrics as well as established measures
from chemoinformatics. As our model is able to capture the complex relationship
between 3d geometry and electronic properties, we bias the distribution of the
generator towards molecules with a small HOMO-LUMO gap - an important property
for the design of organic solar cells.
| stat |
Nested Group Testing Procedures for Screening | This article reviews a class of adaptive group testing procedures that
operate under a probabilistic model assumption as follows. Consider a set of
$N$ items, where item $i$ has the probability $p$ ($p_i$ in the generalized
group testing) to be defective, and the probability $1-p$ to be non-defective
independent from the other items. A group test applied to any subset of size
$n$ is a binary test with two possible outcomes, positive or negative. The
outcome is negative if all $n$ items are non-defective, whereas the outcome is
positive if at least one item among the $n$ items is defective. The goal is
complete identification of all $N$ items with the minimum expected number of
tests.
| stat |
New Development of Bayesian Variable Selection Criteria for Spatial
Point Process with Applications | Selecting important spatial-dependent variables under the nonhomogeneous
spatial Poisson process model is an important topic of great current interest.
In this paper, we use the Deviance Information Criterion (DIC) and Logarithm of
the Pseudo Marginal Likelihood (LPML) for Bayesian variable selection under the
nonhomogeneous spatial Poisson process model. We further derive the new Monte
Carlo estimation formula for LPML in the spatial Poisson process setting.
Extensive simulation studies are carried out to evaluate the empirical
performance of the proposed criteria. The proposed methodology is further
applied to the analysis of two large data sets, the Earthquake Hazards Program
of United States Geological Survey (USGS) earthquake data and the Forest of
Barro Colorado Island (BCI) data.
| stat |
Empirical study towards understanding line search approximations for
training neural networks | Choosing appropriate step sizes is critical for reducing the computational
cost of training large-scale neural network models. Mini-batch sub-sampling
(MBSS) is often employed for computational tractability. However, MBSS
introduces a sampling error, that can manifest as a bias or variance in a line
search. This is because MBSS can be performed statically, where the mini-batch
is updated only when the search direction changes, or dynamically, where the
mini-batch is updated every-time the function is evaluated. Static MBSS results
in a smooth loss function along a search direction, reflecting low variance but
large bias in the estimated "true" (or full batch) minimum. Conversely, dynamic
MBSS results in a point-wise discontinuous function, with computable gradients
using backpropagation, along a search direction, reflecting high variance but
lower bias in the estimated "true" (or full batch) minimum. In this study,
quadratic line search approximations are considered to study the quality of
function and derivative information to construct approximations for dynamic
MBSS loss functions. An empirical study is conducted where function and
derivative information are enforced in various ways for the quadratic
approximations. The results for various neural network problems show that being
selective on what information is enforced helps to reduce the variance of
predicted step sizes.
| stat |
Quantifying the Privacy Risks of Learning High-Dimensional Graphical
Models | Models leak information about their training data. This enables attackers to
infer sensitive information about their training sets, notably determine if a
data sample was part of the model's training set. The existing works
empirically show the possibility of these membership inference (tracing)
attacks against complex deep learning models. However, the attack results are
dependent on the specific training data, can be obtained only after the tedious
process of training the model and performing the attack, and are missing any
measure of the confidence and unused potential power of the attack.
In this paper, we theoretically analyze the maximum power of tracing attacks
against high-dimensional graphical models, with the focus on Bayesian networks.
We provide a tight upper bound on the power (true positive rate) of these
attacks, with respect to their error (false positive rate), for a given model
structure even before learning its parameters. As it should be, the bound is
independent of the knowledge and algorithm of any specific attack. It can help
in identifying which model structures leak more information, how adding new
parameters to the model increases its privacy risk, and what can be gained by
adding new data points to decrease the overall information leakage. It provides
a measure of the potential leakage of a model given its structure, as a
function of the model complexity and the size of the training set.
| stat |
The Nu Class of Low-Degree-Truncated, Rational, Generalized Functions.
Ib. Integrals of Matern-correlation functions for all odd-half-integer class
parameters | This paper is an extension of Parts I and Ia of a series about Nu-class
generalized functions. We provide hand-generated algebraic expressions for
integrals of single Matern-covariance functions, as well as for products of two
Matern-covariance functions, for all odd-half-integer class parameters. These
are useful both for IMSPE-optimal design software and for testing universality
of Nu-class generalized-function properties, across covariance classes.
| stat |
Testing the Equality of Proportions for Combined Unilateral and
Bilateral Data | Measurements are generally collected as unilateral or bilateral data in
clinical trials or observational studies. For example, in ophthalmologic
studies, statistical tests are often based on one or two eyes of an individual.
For bilateral data, recent literatures have shown some testing procedures that
take into account the intra-class correlation between two eyes of the same
person. Ma et al. (2015) investigated three testing procedures under Rosner's
model. In this paper, we extend Ma's work for bilateral data to combined
bilateral and unilateral data. The proposed procedures are based on the
likelihood estimate algorithm derived from the root of 4th order polynomial
equations and fisher scoring iterations. Simulation studies are performed to
compare the testing procedures under different parameter configurations. The
result shows that score test has satisfactory type I error rates and powers.
Therefore, we recommend score test for testing the equality of proportions. We
illustrate the application of the proposed methods with a double-blind
randomized clinical trial.
| stat |
A Statistical Learning Approach to Modal Regression | This paper studies the nonparametric modal regression problem systematically
from a statistical learning view. Originally motivated by pursuing a
theoretical understanding of the maximum correntropy criterion based regression
(MCCR), our study reveals that MCCR with a tending-to-zero scale parameter is
essentially modal regression. We show that nonparametric modal regression
problem can be approached via the classical empirical risk minimization. Some
efforts are then made to develop a framework for analyzing and implementing
modal regression. For instance, the modal regression function is described, the
modal regression risk is defined explicitly and its \textit{Bayes} rule is
characterized; for the sake of computational tractability, the surrogate modal
regression risk, which is termed as the generalization risk in our study, is
introduced. On the theoretical side, the excess modal regression risk, the
excess generalization risk, the function estimation error, and the relations
among the above three quantities are studied rigorously. It turns out that
under mild conditions, function estimation consistency and convergence may be
pursued in modal regression as in vanilla regression protocols, such as mean
regression, median regression, and quantile regression. However, it outperforms
these regression models in terms of robustness as shown in our study from a
re-descending M-estimation view. This coincides with and in return explains the
merits of MCCR on robustness. On the practical side, the implementation issues
of modal regression including the computational algorithm and the tuning
parameters selection are discussed. Numerical assessments on modal regression
are also conducted to verify our findings empirically.
| stat |
Adversarial Noise Attacks of Deep Learning Architectures -- Stability
Analysis via Sparse Modeled Signals | Despite their impressive performance, deep convolutional neural networks
(CNNs) have been shown to be sensitive to small adversarial perturbations.
These nuisances, which one can barely notice, are powerful enough to fool
sophisticated and well performing classifiers, leading to ridiculous
misclassification results. In this paper we analyze the stability of
state-of-the-art deep-learning classification machines to adversarial
perturbations, where we assume that the signals belong to the (possibly
multi-layer) sparse representation model. We start with convolutional sparsity
and then proceed to its multi-layered version, which is tightly connected to
CNNs. Our analysis links between the stability of the classification to noise
and the underlying structure of the signal, quantified by the sparsity of its
representation under a fixed dictionary. In addition, we offer similar
stability theorems for two practical pursuit algorithms, which are posed as two
different deep-learning architectures - the layered Thresholding and the
layered Basis Pursuit. Our analysis establishes the better robustness of the
later to adversarial attacks. We corroborate these theoretical results by
numerical experiments on three datasets: MNIST, CIFAR-10 and CIFAR-100.
| stat |
Comparative Study of Differentially Private Synthetic Data Algorithms
from the NIST PSCR Differential Privacy Synthetic Data Challenge | Differentially private synthetic data generation offers a recent solution to
release analytically useful data while preserving the privacy of individuals in
the data. In order to utilize these algorithms for public policy decisions,
policymakers need an accurate understanding of these algorithms' comparative
performance. Correspondingly, data practitioners require standard metrics for
evaluating the analytic qualities of the synthetic data. In this paper, we
present an in-depth evaluation of several differentially private synthetic data
algorithms using actual differentially private synthetic data sets created by
contestants in the 2018-2019 National Institute of Standards and Technology
Public Safety Communications Research (NIST PSCR) Division's ``Differential
Privacy Synthetic Data Challenge.'' We offer analyses of these algorithms based
on both the accuracy of the data they created and their usability by potential
data providers. We frame the methods used in the NIST PSCR data challenge
within the broader differentially private synthetic data literature. We
implement additional utility metrics, including two of our own, on the
differentially private synthetic data and compare mechanism utility on three
categories. Our comparative assessment of the differentially private data
synthesis methods and the quality metrics shows the relative usefulness, the
general strengths and weaknesses, and offers preferred choices of algorithms
and metrics. Finally we describe the implications of our evaluation for
policymakers seeking to implement differentially private synthetic data
algorithms on future data products.
| stat |
Tractable Post-Selection Maximum Likelihood Inference for the Lasso | Applying standard statistical methods after model selection may yield
inefficient estimators and hypothesis tests that fail to achieve nominal type-I
error rates. The main issue is the fact that the post-selection distribution of
the data differs from the original distribution. In particular, the observed
data is constrained to lie in a subset of the original sample space that is
determined by the selected model. This often makes the post-selection
likelihood of the observed data intractable and maximum likelihood inference
difficult. In this work, we get around the intractable likelihood by generating
noisy unbiased estimates of the post-selection score function and using them in
a stochastic ascent algorithm that yields correct post-selection maximum
likelihood estimates. We apply the proposed technique to the problem of
estimating linear models selected by the lasso. In an asymptotic analysis the
resulting estimates are shown to be consistent for the selected parameters and
to have a limiting truncated normal distribution. Confidence intervals
constructed based on the asymptotic distribution obtain close to nominal
coverage rates in all simulation settings considered, and the point estimates
are shown to be superior to the lasso estimates when the true model is sparse.
| stat |
On Selecting Stable Predictors in Time Series Models | We extend the feature selection methodology to dependent data and propose a
novel time series predictor selection scheme that accommodates statistical
dependence in a more typical i.i.d sub-sampling based framework. Furthermore,
the machinery of mixing stationary processes allows us to quantify the
improvements of our approach over any base predictor selection method (such as
lasso) even in a finite sample setting. Using the lasso as a base procedure we
demonstrate the applicability of our methods to simulated and several real time
series datasets.
| stat |
Adaptive Bayesian Spectral Analysis of High-dimensional Nonstationary
Time Series | This article introduces a nonparametric approach to spectral analysis of a
high-dimensional multivariate nonstationary time series. The procedure is based
on a novel frequency-domain factor model that provides a flexible yet
parsimonious representation of spectral matrices from a large number of
simultaneously observed time series. Real and imaginary parts of the factor
loading matrices are modeled independently using a prior that is formulated
from the tensor product of penalized splines and multiplicative gamma process
shrinkage priors, allowing for infinitely many factors with loadings
increasingly shrunk towards zero as the column index increases. Formulated in a
fully Bayesian framework, the time series is adaptively partitioned into
approximately stationary segments, where both the number and location of
partition points are assumed unknown. Stochastic approximation Monte Carlo
(SAMC) techniques are used to accommodate the unknown number of segments, and a
conditional Whittle likelihood-based Gibbs sampler is developed for efficient
sampling within segments. By averaging over the distribution of partitions, the
proposed method can approximate both abrupt and slowly varying changes in
spectral matrices. Performance of the proposed model is evaluated by extensive
simulations and demonstrated through the analysis of high-density
electroencephalography.
| stat |
Knot Selection in Sparse Gaussian Processes with a Variational Objective
Function | Sparse, knot-based Gaussian processes have enjoyed considerable success as
scalable approximations to full Gaussian processes. Certain sparse models can
be derived through specific variational approximations to the true posterior,
and knots can be selected to minimize the Kullback-Leibler divergence between
the approximate and true posterior. While this has been a successful approach,
simultaneous optimization of knots can be slow due to the number of parameters
being optimized. Furthermore, there have been few proposed methods for
selecting the number of knots, and no experimental results exist in the
literature. We propose using a one-at-a-time knot selection algorithm based on
Bayesian optimization to select the number and locations of knots. We showcase
the competitive performance of this method relative to simultaneous
optimization of knots on three benchmark data sets, but at a fraction of the
computational cost.
| stat |
A state-space model for dynamic functional connectivity | Dynamic functional connectivity (DFC) analysis involves measuring correlated
neural activity over time across multiple brain regions. Significant regional
correlations among neural signals, such as those obtained from resting-state
functional magnetic resonance imaging (fMRI), may represent neural circuits
associated with rest. The conventional approach of estimating the correlation
dynamics as a sequence of static correlations from sliding time-windows has
statistical limitations. To address this issue, we propose a multivariate
stochastic volatility model for estimating DFC inspired by recent work in
econometrics research. This model assumes a state-space framework where the
correlation dynamics of a multivariate normal observation sequence is governed
by a positive-definite matrix-variate latent process. Using this statistical
model within a sequential Bayesian estimation framework, we use blood
oxygenation level dependent activity from multiple brain regions to estimate
posterior distributions on the correlation trajectory. We demonstrate the
utility of this DFC estimation framework by analyzing its performance on
simulated data, and by estimating correlation dynamics in resting state fMRI
data from a patient with a disorder of consciousness (DoC). Our work advances
the state-of-the-art in DFC analysis and its principled use in DoC biomarker
exploration.
| stat |
Scalable Multivariate Histograms | We give a distributed variant of an adaptive histogram estimation procedure
previously developed by the first author. The procedure is based on regular
pavings and is known to have numerous appealing statistical and arithmetical
properties. The distributed version makes it possible to process data sets
significantly bigger than previously. We provide prototype implementation under
a permissive license.
| stat |
Spectral Methods for Data Science: A Statistical Perspective | Spectral methods have emerged as a simple yet surprisingly effective approach
for extracting information from massive, noisy and incomplete data. In a
nutshell, spectral methods refer to a collection of algorithms built upon the
eigenvalues (resp. singular values) and eigenvectors (resp. singular vectors)
of some properly designed matrices constructed from data. A diverse array of
applications have been found in machine learning, data science, and signal
processing. Due to their simplicity and effectiveness, spectral methods are not
only used as a stand-alone estimator, but also frequently employed to
initialize other more sophisticated algorithms to improve performance.
While the studies of spectral methods can be traced back to classical matrix
perturbation theory and methods of moments, the past decade has witnessed
tremendous theoretical advances in demystifying their efficacy through the lens
of statistical modeling, with the aid of non-asymptotic random matrix theory.
This monograph aims to present a systematic, comprehensive, yet accessible
introduction to spectral methods from a modern statistical perspective,
highlighting their algorithmic implications in diverse large-scale
applications. In particular, our exposition gravitates around several central
questions that span various applications: how to characterize the sample
efficiency of spectral methods in reaching a target level of statistical
accuracy, and how to assess their stability in the face of random noise,
missing data, and adversarial corruptions? In addition to conventional $\ell_2$
perturbation analysis, we present a systematic $\ell_{\infty}$ and
$\ell_{2,\infty}$ perturbation theory for eigenspace and singular subspaces,
which has only recently become available owing to a powerful "leave-one-out"
analysis framework.
| stat |
Certifying Some Distributional Robustness with Principled Adversarial
Training | Neural networks are vulnerable to adversarial examples and researchers have
proposed many heuristic attack and defense mechanisms. We address this problem
through the principled lens of distributionally robust optimization, which
guarantees performance under adversarial input perturbations. By considering a
Lagrangian penalty formulation of perturbing the underlying data distribution
in a Wasserstein ball, we provide a training procedure that augments model
parameter updates with worst-case perturbations of training data. For smooth
losses, our procedure provably achieves moderate levels of robustness with
little computational or statistical cost relative to empirical risk
minimization. Furthermore, our statistical guarantees allow us to efficiently
certify robustness for the population loss. For imperceptible perturbations,
our method matches or outperforms heuristic approaches.
| stat |
Near-Linear Time Local Polynomial Nonparametric Estimation with Box
Kernels | Local polynomial regression (Fan and Gijbels 1996) is an important class of
methods for nonparametric density estimation and regression problems. However,
straightforward implementation of local polynomial regression has quadratic
time complexity which hinders its applicability in large-scale data analysis.
In this paper, we significantly accelerate the computation of local polynomial
estimates by novel applications of multi-dimensional binary indexed trees
(Fenwick 1994). Both time and space complexity of our proposed algorithm is
nearly linear in the number of input data points. Simulation results confirm
the efficiency and effectiveness of our proposed approach.
| stat |
Bayesian influence diagnostics and outlier detection for meta-analysis
of diagnostic test accuracy | Meta-analyses of diagnostic test accuracy (DTA) studies have been gaining
prominence in research in clinical epidemiology and health technology
development. In these DTA meta-analyses, some studies may have markedly
different characteristics from the others, and potentially be inappropriate to
include. The inclusion of these "outlying" studies might lead to biases,
yielding misleading results. In addition, there might be influential studies
that have notable impacts on the results. In this article, we propose Bayesian
methods for detecting outlying studies and their influence diagnostics in DTA
meta-analyses. Synthetic influence measures based on the bivariate hierarchical
Bayesian random effects models are developed because the overall influences of
individual studies should be simultaneously assessed by the two outcome
variables and their correlation information. We propose four synthetic measures
for influence analyses: (1) relative distance, (2) standardized residual, (3)
Bayesian p-value, and (4) influence statistic on the area under the summary
receiver operating characteristic curve. We also show conventional univariate
Bayesian influential measures can be applied to the bivariate random effects
models, which can be used as marginal influential measures. We illustrate the
effectiveness of the proposed methods by applying them to a DTA meta-analysis
of ultrasound in screening for vesicoureteral reflux among children with
urinary tract infections.
| stat |
Think-aloud interviews: A tool for exploring student statistical
reasoning | As undergraduate statistics education rapidly changes to incorporate new
topics and skills, statistics educators need tools to evaluate student learning
and understand how students think. Think-aloud interviews, in which students
answer questions while narrating their thinking aloud, are a valuable education
research tool for detecting misconceptions and developing robust assessments.
While think-aloud interviews have been widely used for education research in
other fields, in statistics education they have been primarily used to vet
concept inventory questions for confusing wording or poor design. Here, we
argue for their much more comprehensive use to explore student reasoning and
iteratively draft questions. To motivate the use of think-aloud interviews, we
describe two case studies on correlation, causation, and sampling
distributions. In these, think-alouds revealed unexpected student reasoning and
suggested new ways to assess difficult statistical topics. These case studies
illustrate the usefulness of think-aloud interviews for studying student
thinking, and we argue for their wider use in statistics education research.
| stat |
Generalization Bounds in the Presence of Outliers: a Median-of-Means
Study | In contrast to the empirical mean, the Median-of-Means (MoM) is an estimator
of the mean $\theta$ of a square integrable r.v. $Z$, around which accurate
nonasymptotic confidence bounds can be built, even when $Z$ does not exhibit a
sub-Gaussian tail behavior. Thanks to the high confidence it achieves on
heavy-tailed data, MoM has found various applications in machine learning,
where it is used to design training procedures that are not sensitive to
atypical observations. More recently, a new line of work is now trying to
characterize and leverage MoM's ability to deal with corrupted data. In this
context, the present work proposes a general study of MoM's concentration
properties under the contamination regime, that provides a clear understanding
of the impact of the outlier proportion and the number of blocks chosen. The
analysis is extended to (multisample) $U$-statistics, i.e. averages over tuples
of observations, that raise additional challenges due to the dependence
induced. Finally, we show that the latter bounds can be used in a
straightforward fashion to derive generalization guarantees for pairwise
learning in a contaminated setting, and propose an algorithm to compute
provably reliable decision functions.
| stat |
Network Analysis of SFU Course Registrations | We investigate the dynamics of disease infection via shared classes at Simon
Fraser University, a medium-sized school in Western Canada. Specifically, we
use simulation to examine the impact of keeping classes above a certain size
online, while those below that size are allowed to meet in person. We use
simple models for infection and recovery, as well as multiple regimes for
infectiousness and class size threshold. Graph theoretic properties of the
student-class enrollment network are also computed, and one such statistic is
used to model an important output of our simulations.
| stat |
Optimistic bounds for multi-output prediction | We investigate the challenge of multi-output learning, where the goal is to
learn a vector-valued function based on a supervised data set. This includes a
range of important problems in Machine Learning including multi-target
regression, multi-class classification and multi-label classification. We begin
our analysis by introducing the self-bounding Lipschitz condition for
multi-output loss functions, which interpolates continuously between a
classical Lipschitz condition and a multi-dimensional analogue of a smoothness
condition. We then show that the self-bounding Lipschitz condition gives rise
to optimistic bounds for multi-output learning, which are minimax optimal up to
logarithmic factors. The proof exploits local Rademacher complexity combined
with a powerful minoration inequality due to Srebro, Sridharan and Tewari. As
an application we derive a state-of-the-art generalization bound for
multi-class gradient boosting.
| stat |
Hamiltonian MCMC methods for estimating rare events probabilities in
high-dimensional problems | Accurate and efficient estimation of rare events probabilities is of
significant importance, since often the occurrences of such events have
widespread impacts. The focus in this work is on precisely quantifying these
probabilities, often encountered in reliability analysis of complex engineering
systems, based on an introduced framework termed Approximate Sampling Target
with Post-processing Adjustment (ASTPA), which herein is integrated with and
supported by gradient-based Hamiltonian Markov Chain Monte Carlo (HMCMC)
methods. The basic idea is to construct a relevant target distribution by
weighting the high-dimensional random variable space through a one-dimensional
output likelihood model, using the limit-state function. To sample from this
target distribution, we exploit HMCMC algorithms, a family of MCMC methods that
adopts physical system dynamics, rather than solely using a proposal
probability distribution, to generate distant sequential samples, and we
develop a new Quasi-Newton mass preconditioned HMCMC scheme (QNp-HMCMC), which
is particularly efficient and suitable for high-dimensional spaces. To
eventually compute the rare event probability, an original post-sampling step
is devised using an inverse importance sampling procedure based on the already
obtained samples. The statistical properties of the estimator are analyzed as
well, and the performance of the proposed methodology is examined in detail and
compared against Subset Simulation in a series of challenging low- and
high-dimensional problems.
| stat |
Monitoring the COVID-19 epidemic with nationwide telecommunication data | In response to the novel coronavirus disease (COVID-19), governments have
introduced severe policy measures with substantial effects on human behavior.
Here, we perform a large-scale, spatio-temporal analysis of human mobility
during the COVID-19 epidemic. We derive human mobility from anonymized,
aggregated telecommunication data in a nationwide setting (Switzerland;
February 10 - April 26, 2020), consisting of ~1.5 billion trips. In comparison
to the same time period from 2019, human movement in Switzerland dropped by
49.1%. The strongest reduction is linked to bans on gatherings of more than 5
people, which is estimated to have decreased mobility by 24.9%, followed by
venue closures (stores, restaurants, and bars) and school closures. As such,
human mobility at a given day predicts reported cases 7-13 days ahead. A 1%
reduction in human mobility predicts a 0.88-1.11% reduction in daily reported
COVID-19 cases. When managing epidemics, monitoring human mobility via
telecommunication data can support public decision-makers in two ways. First,
it helps in assessing policy impact; second, it provides a scalable tool for
near real-time epidemic surveillance, thereby enabling evidence-based policies.
| stat |
Risk models for breast cancer and their validation | Strategies to prevent cancer and diagnose it early when it is most treatable
are needed to reduce the public health burden from rising disease incidence.
Risk assessment is playing an increasingly important role in targeting
individuals in need of such interventions. For breast cancer many individual
risk factors have been well understood for a long time, but the development of
a fully comprehensive risk model has not been straightforward, in part because
there have been limited data where joint effects of an extensive set of risk
factors may be estimated with precision. In this article we first review the
approach taken to develop the IBIS (Tyrer-Cuzick) model, and describe recent
updates. We then review and develop methods to assess calibration of models
such as this one, where the risk of disease allowing for competing mortality
over a long follow-up time or lifetime is estimated. The breast cancer risk
model model and calibration assessment methods are demonstrated using a cohort
of 132 139 women attending mammography screening in Washington, USA.
| stat |
Self-Bounding Majority Vote Learning Algorithms by the Direct
Minimization of a Tight PAC-Bayesian C-Bound | In the PAC-Bayesian literature, the C-Bound refers to an insightful relation
between the risk of a majority vote classifier (under the zero-one loss) and
the first two moments of its margin (i.e., the expected margin and the voters'
diversity). Until now, learning algorithms developed in this framework minimize
the empirical version of the C-Bound, instead of explicit PAC-Bayesian
generalization bounds. In this paper, by directly optimizing PAC-Bayesian
guarantees on the C-Bound, we derive self-bounding majority vote learning
algorithms. Moreover, our algorithms based on gradient descent are scalable and
lead to accurate predictors paired with non-vacuous guarantees.
| stat |
Automated Scalable Bayesian Inference via Hilbert Coresets | The automation of posterior inference in Bayesian data analysis has enabled
experts and nonexperts alike to use more sophisticated models, engage in faster
exploratory modeling and analysis, and ensure experimental reproducibility.
However, standard automated posterior inference algorithms are not tractable at
the scale of massive modern datasets, and modifications to make them so are
typically model-specific, require expert tuning, and can break theoretical
guarantees on inferential quality. Building on the Bayesian coresets framework,
this work instead takes advantage of data redundancy to shrink the dataset
itself as a preprocessing step, providing fully-automated, scalable Bayesian
inference with theoretical guarantees. We begin with an intuitive reformulation
of Bayesian coreset construction as sparse vector sum approximation, and
demonstrate that its automation and performance-based shortcomings arise from
the use of the supremum norm. To address these shortcomings we develop Hilbert
coresets, i.e., Bayesian coresets constructed under a norm induced by an
inner-product on the log-likelihood function space. We propose two Hilbert
coreset construction algorithms---one based on importance sampling, and one
based on the Frank-Wolfe algorithm---along with theoretical guarantees on
approximation quality as a function of coreset size. Since the exact
computation of the proposed inner-products is model-specific, we automate the
construction with a random finite-dimensional projection of the log-likelihood
functions. The resulting automated coreset construction algorithm is simple to
implement, and experiments on a variety of models with real and synthetic
datasets show that it provides high-quality posterior approximations and a
significant reduction in the computational cost of inference.
| stat |
New Results for Pearson Type III Family of Distributions and Application
in Wireless Power Transfer | Pearson and log Pearson type III distributions have been considered in
several scientific fields, as in hydrology and seismology. In this paper, we
present new results for these distributions and we utilize them, for first time
in the literature, to investigate the statistical behavior of wireless power
transfer (WPT), assuming that the harvested energy follows a well-established
nonlinear energy harvesting model based on the logistic function. Specifically,
we present new closed-form expressions for the statistical properties of a
general form of Pearson and log Pearson type III distributions and we utilize
them to introduce a new member of the Pearson type III family, the logit
Pearson type III distribution, through which the logit gamma distribution is
also defined. Moreover, we derive closed-form expressions for the probability
density function, the cumulative distribution function and moments of the
distributions of the sum, the log sum and the logit sum of Pearson type III
random variables. Furthermore, taking into account that Pearson type III family
of distributions is closely related to the considered nonlinear harvesting
model the statistical properties of the distribution of the harvested power and
derived, for both single input single output and multiple input single output
scenarios.
| stat |
Modeling Social Media User Content Generation Using Interpretable Point
Process Models | In this article, we study the activity patterns of modern social media users
on platforms such as Twitter and Facebook. To characterize the complex patterns
we observe in users' interactions with social media, we describe a new class of
point process models. The components in the model have straightforward
interpretations and can thus provide meaningful insights into user activity
patterns. A composite likelihood approach and a composite EM estimation
procedure are developed to overcome the challenges that arise in parameter
estimation. Using the proposed method, we analyze Donald Trump's Twitter data
and study if and how his tweeting behavior evolved before, during and after the
presidential campaign. Additionally, we analyze a large-scale social media data
from Sina Weibo and identify interesting groups of users with distinct
behaviors; in this analysis, we also discuss the effect of social ties on a
user's online content generating behavior.
| stat |
Bayesian non-parametric ordinal regression under a monotonicity
constraint | Compared to the nominal scale, ordinal scale for a categorical outcome
variable has the property of making a monotonicity assumption for the covariate
effects meaningful. This assumption is encoded in the commonly used
proportional odds model, but there it is combined with other parametric
assumptions such as linearity and additivity. Herein, we consider models where
monotonicity is used as the only modeling assumption when modeling the effects
of covariates on the cumulative probabilities of ordered outcome categories. We
are not aware of other non-parametric multivariable monotonic ordinal models
for inference purposes. We generalize our previously proposed Bayesian
monotonic multivariable regression model to ordinal outcomes, and propose an
estimation procedure based on reversible jump Markov chain Monte Carlo. The
model is based on a marked point process construction, which allows it to
approximate arbitrary monotonic regression function shapes, and has a built-in
covariate selection property. We study the performance of the proposed approach
through extensive simulation studies, and demonstrate its practical application
in two real data examples.
| stat |
Bayesian inference for dynamic vine copulas in higher dimensions | We propose a class of dynamic vine copula models. This is an extension of
static vine copulas and a generalization of dynamic C-vine and D-vine copulas
studied by Almeida et al (2016) and Goel and Mehra (2019). Within this class,
we allow for time-varying dependence by driving the vine copula parameters with
latent AR(1) processes. This modeling approach is very flexible but estimation
is not straightforward due to the high-dimensional parameter space. We propose
a Bayesian estimation approach, which relies on a novel approximation of the
posterior distribution. This approximation allows to use Markov Chain Monte
Carlo methods, such as elliptical slice sampling, in a sequential way. In
contrast to other Bayesian sequential estimation procedures for vine copula
models as proposed by Gruber and Czado (2015), there is no need to collapse
copula parameters to point estimates before proceeding to the next tree. Thus
more information and uncertainty is propagated from lower to higher trees. A
simulation study shows satisfactory performance of the Bayesian procedure. This
dynamic modeling and inference approach can be applied in various fields, where
static vine copulas have already proven to be successful, including
environmental sciences, medicine and finance. Here we study the dependence
among 21 exchange rates. For comparison we also estimate a static vine copula
model and dynamic C-vine and D-vine copula models. This comparison shows
superior performance of the proposed dynamic vine copula model with respect to
one day ahead forecasting accuracy.
| stat |
A General Method for Robust Learning from Batches | In many applications, data is collected in batches, some of which are corrupt
or even adversarial. Recent work derived optimal robust algorithms for
estimating discrete distributions in this setting. We consider a general
framework of robust learning from batches, and determine the limits of both
classification and distribution estimation over arbitrary, including
continuous, domains. Building on these results, we derive the first robust
agnostic computationally-efficient learning algorithms for piecewise-interval
classification, and for piecewise-polynomial, monotone, log-concave, and
gaussian-mixture distribution estimation.
| stat |
Bayesian Analysis of Privacy Attacks on GPS Trajectories | The success of applications for sharing GPS trajectories raises serious
privacy concerns, in particular about users' home addresses. In this paper we
show that a Bayesian approach is natural and effective for a rigorous analysis
of home-identification attacks and their countermeasures, in terms of privacy.
We focus on a family of countermeasures named "privacy-region strategies",
consisting in publishing each trajectory from the first exit to the last
entrance from/into a privacy region. Their performance is studied through
simulations on Brownian motions.
| stat |
The Information Content of Taster's Valuation in Tea Auctions of India | Tea auctions across India occur as an ascending open auction, conducted
online. Before the auction, a sample of the tea lot is sent to potential
bidders and a group of tea tasters. The seller's reserve price is a
confidential function of the tea taster's valuation, which also possibly acts
as a signal to the bidders.
In this paper, we work with the dataset from a single tea auction house, J
Thomas, of tea dust category, on 49 weeks in the time span of 2018-2019, with
the following objectives in mind:
$\bullet$ Objective classification of the various categories of tea dust (25)
into a more manageable, and robust classification of the tea dust, based on
source and grades.
$\bullet$ Predict which tea lots would be sold in the auction market, and a
model for the final price conditioned on sale.
$\bullet$ To study the distribution of price and ratio of the sold tea
auction lots.
$\bullet$ Make a detailed analysis of the information obtained from the tea
taster's valuation and its impact on the final auction price.
The model used has shown various promising results on cross-validation. The
importance of valuation is firmly established through analysis of causal
relationship between the valuation and the actual price. The authors hope that
this study of the properties and the detailed analysis of the role played by
the various factors, would be significant in the decision making process for
the players of the auction game, pave the way to remove the manual interference
in an attempt to automate the auction procedure, and improve tea quality in
markets.
| stat |
Region-Referenced Spectral Power Dynamics of EEG Signals: A Hierarchical
Modeling Approach | Functional brain imaging through electroencephalography (EEG) relies upon the
analysis and interpretation of high-dimensional, spatially organized time
series. We propose to represent time-localized frequency domain
characterizations of EEG data as region-referenced functional data. This
representation is coupled with a hierarchical modeling approach to multivariate
functional observations. Within this familiar setting, we discuss how several
prior models relate to structural assumptions about multivariate covariance
operators. An overarching modeling framework, based on infinite factorial
decompositions, is finally proposed to balance flexibility and efficiency in
estimation. The motivating application stems from a study of implicit auditory
learning, in which typically developing (TD) children, and children with autism
spectrum disorder (ASD) were exposed to a continuous speech stream. Using the
proposed model, we examine differential band power dynamics as brain function
is interrogated throughout the duration of a computer-controlled experiment.
Our work offers a novel look at previous findings in psychiatry, and provides
further insights into the understanding of ASD. Our approach to inference is
fully Bayesian and implemented in a highly optimized Rcpp package.
| stat |
USP: an independence test that improves on Pearson's chi-squared and the
$G$-test | We present the $U$-Statistic Permutation (USP) test of independence in the
context of discrete data displayed in a contingency table. Either Pearson's
chi-squared test of independence, or the $G$-test, are typically used for this
task, but we argue that these tests have serious deficiencies, both in terms of
their inability to control the size of the test, and their power properties. By
contrast, the USP test is guaranteed to control the size of the test at the
nominal level for all sample sizes, has no issues with small (or zero) cell
counts, and is able to detect distributions that violate independence in only a
minimal way. The test statistic is derived from a $U$-statistic estimator of a
natural population measure of dependence, and we prove that this is the unique
minimum variance unbiased estimator of this population quantity. The practical
utility of the USP test is demonstrated on both simulated data, where its power
can be dramatically greater than those of Pearson's test and the $G$-test, and
on real data. The USP test is implemented in the R package USP.
| stat |
An Evaluation of Design-based Properties of Different Composite
Estimators | For the last several decades, the US Census Bureau has been using the AK
composite estimation method to produce statistics on employment from the
Current Population Survey (CPS) data. The CPS uses a rotating design and AK
estimators are linear combinations of monthly survey weighted averages (called
month-in-sample estimates) in each rotation groups. Denoting by $X$ the vector
of month-in-sample estimates and by $\Sigma$ its design based variance, the
coefficients of the linear combination were optimized by the Census Bureau
after substituting $\Sigma$ by an estimate and under unrealistic stationarity
assumptions. To show the limits of this approach, we compared the AK estimator
with different competitors using three different synthetic populations that
mimics the Current Population Survey (CPS) data and a simplified sample design
that mimics the CPS design. In our simulation setup, empirically best
estimators have larger mean square error than simple averages. In the real data
analysis, the AK estimates are constantly below the survey-weighted estimates,
indicating potential bias. Any attempt to improve on the estimated optimal
estimator in either class would require a thorough investigation of the highly
non-trivial problem of estimation of $\Sigma$ for a complex setting like the
CPS (we did not entertain this problem in this paper). A different approach is
to use a variant of the regression composite estimator used by Statistics
Canada. The regression composite estimator does not require estimation of
$\Sigma$ and is less sensitive to the rotation group bias in our simulations.
Our study demonstrates that there is a great potential for improving the
estimation of levels and month to month changes in the unemployment rates by
using the regression composite estimator.
| stat |
Estimation and uncertainty quantification for extreme quantile regions | Estimation of extreme quantile regions, spaces in which future extreme events
can occur with a given low probability, even beyond the range of the observed
data, is an important task in the analysis of extremes. Existing methods to
estimate such regions are available, but do not provide any measures of
estimation uncertainty. We develop univariate and bivariate schemes for
estimating extreme quantile regions under the Bayesian paradigm that
outperforms existing approaches and provides natural measures of quantile
region estimate uncertainty. We examine the method's performance in controlled
simulation studies. We illustrate the applicability of the proposed method by
analysing high bivariate quantiles for pairs of pollutants, conditionally on
different temperature gradations, recorded in Milan, Italy.
| stat |
Study of the impact of climate change on precipitation in Paris area
using method based on iterative multiscale dynamic time warping (IMS-DTW) | Studying the impact of climate change on precipitation is constrained by
finding a way to evaluate the evolution of precipitation variability over time.
Classical approaches (feature-based) have shown their limitations for this
issue due to the intermittent and irregular nature of precipitation. In this
study, we present a novel variant of the Dynamic time warping method
quantifying the dissimilarity between two rainfall time series based on shapes
comparisons, for clustering annual time series recorded at daily scale. This
shape based approach considers the whole information (variability, trends and
intermittency). We further labeled each cluster using a feature-based approach.
While testing the proposed approach on the time series of Paris Montsouris, we
found that the precipitation variability increased over the years in Paris
area.
| stat |
Bayesian Uncertainty Quantification for Low-rank Matrix Completion | We consider the problem of uncertainty quantification for an unknown low-rank
matrix $\mathbf{X}$, given a partial and noisy observation of its entries. This
quantification of uncertainty is essential for many real-world problems,
including image processing, satellite imaging, and seismology, providing a
principled framework for validating scientific conclusions and guiding
decision-making. However, existing literature has largely focused on the
completion (i.e., point estimation) of the matrix $\mathbf{X}$, with little
work on investigating its uncertainty. To this end, we propose in this work a
new Bayesian modeling framework, called BayeSMG, which parametrizes the unknown
$\mathbf{X}$ via its underlying row and column subspaces. This Bayesian
subspace parametrization allows for efficient posterior inference on matrix
subspaces, which represents interpretable phenomena in many applications. This
can then be leveraged for improved matrix recovery. We demonstrate the
effectiveness of BayeSMG over existing Bayesian matrix recovery methods in
numerical experiments and a seismic sensor network application.
| stat |
Learning stable and predictive structures in kinetic systems: Benefits
of a causal approach | Learning kinetic systems from data is one of the core challenges in many
fields. Identifying stable models is essential for the generalization
capabilities of data-driven inference. We introduce a computationally efficient
framework, called CausalKinetiX, that identifies structure from discrete time,
noisy observations, generated from heterogeneous experiments. The algorithm
assumes the existence of an underlying, invariant kinetic model, a key
criterion for reproducible research. Results on both simulated and real-world
examples suggest that learning the structure of kinetic systems benefits from a
causal perspective. The identified variables and models allow for a concise
description of the dynamics across multiple experimental settings and can be
used for prediction in unseen experiments. We observe significant improvements
compared to well established approaches focusing solely on predictive
performance, especially for out-of-sample generalization.
| stat |
Warp Bridge Sampling: The Next Generation | Bridge sampling is an effective Monte Carlo method for estimating the ratio
of normalizing constants of two probability densities, a routine computational
problem in statistics, physics, chemistry, and other fields. The Monte Carlo
error of the bridge sampling estimator is determined by the amount of overlap
between the two densities. In the case of uni-modal densities, Warp-I, II, and
III transformations (Meng and Schilling, 2002) are effective for increasing the
initial overlap, but they are less so for multi-modal densities. This paper
introduces Warp-U transformations that aim to transform multi-modal densities
into Uni-modal ones without altering their normalizing constants. The
construction of a Warp-U transformation starts with a Normal (or other
convenient) mixture distribution $\phi_{\text{mix}}$ that has reasonable
overlap with the target density $p$, whose normalizing constant is unknown. The
stochastic transformation that maps $\phi_{\text{mix}}$ back to its generating
distribution $N(0,1)$ is then applied to $p$ yielding its Warp-U version, which
we denote $\tilde{p}$. Typically, $\tilde{p}$ is uni-modal and has
substantially increased overlap with $N(0,1)$. Furthermore, we prove that the
overlap between $\tilde{p}$ and $N(0,1)$ is guaranteed to be no less than the
overlap between $p$ and $\phi_{\text{mix}}$, in terms of any $f$-divergence. We
propose a computationally efficient method to find an appropriate
$\phi_{\text{mix}}$, and a simple but effective approach to remove the bias
which results from estimating the normalizing constants and fitting
$\phi_{\text{mix}}$ with the same data. We illustrate our findings using 10 and
50 dimensional highly irregular multi-modal densities, and demonstrate how
Warp-U sampling can be used to improve the final estimation step of the
Generalized Wang-Landau algorithm (Liang, 2005), a powerful sampling and
estimation method.
| stat |
Integrating Phase 2 into Phase 3 based on an Intermediate Endpoint While
Accounting for a Cure Proportion -- with an Application to the Design of a
Clinical Trial in Acute Myeloid Leukemia | For a trial with primary endpoint overall survival for a molecule with
curative potential, statistical methods that rely on the proportional hazards
assumption may underestimate the power and the time to final analysis. We show
how a cure proportion model can be used to get the necessary number of events
and appropriate timing via simulation. If Phase 1 results for the new drug are
exceptional and/or the medical need in the target population is high, a Phase 3
trial might be initiated after Phase 1. Building in a futility interim analysis
into such a pivotal trial may mitigate the uncertainty of moving directly to
Phase 3. However, if cure is possible, overall survival might not be mature
enough at the interim to support a futility decision. We propose to base this
decision on an intermediate endpoint that is sufficiently associated with
survival. Planning for such an interim can be interpreted as making a
randomized Phase 2 trial a part of the pivotal trial: if stopped at the
interim, the trial data would be analyzed and a decision on a subsequent Phase
3 trial would be made. If the trial continues at the interim then the Phase 3
trial is already underway. To select a futility boundary, a mechanistic
simulation model that connects the intermediate endpoint and survival is
proposed. We illustrate how this approach was used to design a pivotal
randomized trial in acute myeloid leukemia, discuss historical data that
informed the simulation model, and operational challenges when implementing it.
| stat |
Log-Normalization Constant Estimation using the Ensemble Kalman-Bucy
Filter with Application to High-Dimensional Models | In this article we consider the estimation of the log-normalization constant
associated to a class of continuous-time filtering models. In particular, we
consider ensemble Kalman-Bucy filter based estimates based upon several
nonlinear Kalman-Bucy diffusions. Based upon new conditional bias results for
the mean of the afore-mentioned methods, we analyze the empirical log-scale
normalization constants in terms of their $\mathbb{L}_n-$errors and conditional
bias. Depending on the type of nonlinear Kalman-Bucy diffusion, we show that
these are of order $(\sqrt{t/N}) + t/N$ or $1/\sqrt{N}$ ($\mathbb{L}_n-$errors)
and of order $[t+\sqrt{t}]/N$ or $1/N$ (conditional bias), where $t$ is the
time horizon and $N$ is the ensemble size. Finally, we use these results for
online static parameter estimation for above filtering models and implement the
methodology for both linear and nonlinear models.
| stat |
VarGrad: A Low-Variance Gradient Estimator for Variational Inference | We analyse the properties of an unbiased gradient estimator of the ELBO for
variational inference, based on the score function method with leave-one-out
control variates. We show that this gradient estimator can be obtained using a
new loss, defined as the variance of the log-ratio between the exact posterior
and the variational approximation, which we call the $\textit{log-variance
loss}$. Under certain conditions, the gradient of the log-variance loss equals
the gradient of the (negative) ELBO. We show theoretically that this gradient
estimator, which we call $\textit{VarGrad}$ due to its connection to the
log-variance loss, exhibits lower variance than the score function method in
certain settings, and that the leave-one-out control variate coefficients are
close to the optimal ones. We empirically demonstrate that VarGrad offers a
favourable variance versus computation trade-off compared to other
state-of-the-art estimators on a discrete VAE.
| stat |
Optimal Study Design for Reducing Variances of Coefficient Estimators in
Change-Point Models | In longitudinal studies, we observe measurements of the same variables at
different time points to track the changes in their pattern over time. In such
studies, scheduling of the data collection waves (i.e. time of participants'
visits) is often pre-determined to accommodate ease of project management and
compliance. Hence, it is common to schedule those visits at equally spaced time
intervals. However, recent publications based on simulated experiments indicate
that the power of studies and the precision of model parameter estimators is
related to the participants' visiting schemes.
In this paper, we consider the longitudinal studies that investigate the
changing pattern of a disease outcome, (e.g. the accelerated cognitive decline
of senior adults). Such studies are often analyzed by the broken-stick model,
consisting of two segments of linear models connected at an unknown
change-point. We formulate this design problem into a high-dimensional
optimization problem and derive its analytical solution. Based on this
solution, we propose an optimal design of the visiting scheme that maximizes
the power (i.e. reduce the variance of estimators) to identify the onset of
accelerated decline. Using both simulation studies and evidence from real data,
we demonstrate our optimal design outperforms the standard equally-spaced
design.
Applying our novel design to plan the longitudinal studies, researchers can
improve the power of detecting pattern change without collecting extra data.
| stat |
Robust optimal estimation of location from discretely sampled functional
data | Estimating location is a central problem in functional data analysis, yet
most current estimation procedures either unrealistically assume completely
observed trajectories or lack robustness with respect to the many kinds of
anomalies one can encounter in the functional setting. To remedy these
deficiencies we introduce the first class of optimal robust location estimators
based on discretely sampled functional data. The proposed method is based on
M-type smoothing spline estimation with repeated measurements and is suitable
for both commonly and independently observed trajectories that are subject to
measurement error. We show that under suitable assumptions the proposed family
of estimators is minimax rate optimal both for commonly and independently
observed trajectories and we illustrate its highly competitive performance and
practical usefulness in a Monte-Carlo study and a real-data example involving
recent Covid-19 data.
| stat |
An Asympirical Smoothing Parameters Selection Approach for Smoothing
Spline ANOVA Models in Large Samples | Large samples have been generated routinely from various sources. Classic
statistical models, such as smoothing spline ANOVA models, are not well
equipped to analyze such large samples due to expensive computational costs. In
particular, the daunting computational costs of selecting smoothing parameters
render smoothing spline ANOVA models impractical. In this article, we develop
an asympirical, i.e., asymptotic and empirical, smoothing parameters selection
approach for smoothing spline ANOVA models in large samples. The idea of this
approach is to use asymptotic analysis to show that the optimal smoothing
parameter is a polynomial function of the sample size and an unknown constant.
The unknown constant is then estimated through empirical subsample
extrapolation. The proposed method significantly reduces the computational
costs of selecting smoothing parameters in high-dimensional and large samples.
We show smoothing parameters chosen by the proposed method tend to the optimal
smoothing parameters that minimise a specific risk function. In addition, the
estimator based on the proposed smoothing parameters achieves the optimal
convergence rate. Extensive simulation studies demonstrate the numerical
advantage of the proposed method over competing methods in terms of relative
efficacies and running time. On an application to molecular dynamics data with
nearly one million observations, the proposed method has the best prediction
performance.
| stat |
Leveraging vague prior information in general models via iteratively
constructed Gamma-minimax estimators | Gamma-minimax estimation is an approach to incorporate prior information into
an estimation procedure when it is implausible to specify one particular prior
distribution. In this approach, we aim for an estimator that minimizes the
worst-case Bayes risk over a set $\Gamma$ of prior distributions.
Traditionally, Gamma-minimax estimation is defined for parametric models. In
this paper, we define Gamma-minimaxity for general models and propose iterative
algorithms with convergence guarantees to compute Gamma-minimax estimators for
a general model space and a set of prior distributions constrained by
generalized moments. We also propose encoding the space of candidate estimators
by neural networks to enable flexible estimation. We illustrate our method in
two settings, namely entropy estimation and a problem that arises in
biodiversity studies.
| stat |
Feature-wise change detection and robust indoor positioning using
RANSAC-like approach | Fingerprinting-based positioning, one of the promising indoor positioning
solutions, has been broadly explored owing to the pervasiveness of sensor-rich
mobile devices, the prosperity of opportunistically measurable
location-relevant signals and the progress of data-driven algorithms. One
critical challenge is to controland improve the quality of the reference
fingerprint map (RFM), which is built at the offline stage and applied for
online positioning. The key concept concerningthe quality control of the RFM is
updating the RFM according to the newly measured data. Though varies methods
have been proposed for adapting the RFM, they approach the problem by
introducing extra-positioning schemes (e.g. PDR orUGV) and directly adjust the
RFM without distinguishing whether critical changes have occurred. This paper
aims at proposing an extra-positioning-free solution by making full use of the
redundancy of measurable features. Loosely inspired by random sampling
consensus (RANSAC), arbitrarily sampled subset of features from the online
measurement are used for generating multi-resamples, which areused for
estimating the intermediate locations. In the way of resampling, it can
mitigate the impact of the changed features on positioning and enables to
retrieve accurate location estimation. The users location is robustly computed
by identifying the candidate locations from these intermediate ones using
modified Jaccardindex (MJI) and the feature-wise change belief is calculated
according to the world model of the RFM and the estimated variability of
features. In order to validate our proposed approach, two levels of
experimental analysis have been carried out. On the simulated dataset, the
average change detection accuracy is about 90%. Meanwhile, the improvement of
positioning accuracy within 2 m is about 20% by dropping out the features that
are detected as changed when performing positioning comparing to that of using
all measured features for location estimation. On the long-term collected
dataset, the average change detection accuracy is about 85%.
| stat |
Six Years of Shiny in Research -- Collaborative Development of Web Tools
in R | The use of Shiny in research publications is investigated. From the
appearance of this popular web application framework for R through to 2018, it
has been utilised in many diverse research areas. While it can be shown that
the complexity of Shiny applications is limited by the background architecture,
and real security concerns exist for novice app developers, the collaborative
benefits are worth attention from the wider research community. Shiny
simplifies the use of complex methodologies for users of different
specialities, at the level of proficiency appropriate for the end user. This
enables a diverse community of users to interact efficiently, utilising
cutting-edge methodologies. The literature reviewed demonstrates that complex
methodologies can be put into practice without the necessity for investment in
professional training. It would appear that Shiny opens up concurrent benefits
in communication between those who analyse data and those in other disciplines,
thereby potentially enriching research through this technology.
| stat |
Conditional Normal Extreme-Value Copulas | We propose a new class of extreme-value copulas which are extreme-value
limits of conditional normal models. Conditional normal models are
generalizations of conditional independence models, where the dependence among
observed variables is modeled using one unobserved factor. Conditional on this
factor, the distribution of these variables is given by the Gaussian copula.
This structure allows one to build flexible and parsimonious models for data
with complex dependence structures, such as data with spatial dependence or
factor structure. We study the extreme-value limits of these models and show
some interesting special cases of the proposed class of copulas. We develop
estimation methods for the proposed models and conduct a simulation study to
assess the performance of these algorithms. Finally, we apply these copula
models to analyze data on monthly wind maxima and stock return minima.
| stat |
Statistical Analysis of Functions on Surfaces, with an application to
Medical Imaging | In Functional Data Analysis, data are commonly assumed to be smooth functions
on a fixed interval of the real line. In this work, we introduce a
comprehensive framework for the analysis of functional data, whose domain is a
two-dimensional manifold and the domain itself is subject to variability from
sample to sample. We formulate a statistical model for such data, here called
Functions on Surfaces, which enables a joint representation of the geometric
and functional aspects, and propose an associated estimation framework. We
assess the validity of the framework by performing a simulation study and we
finally apply it to the analysis of neuroimaging data of cortical thickness,
acquired from the brains of different subjects, and thus lying on domains with
different geometries.
| stat |
A Theorem of the Alternative for Personalized Federated Learning | A widely recognized difficulty in federated learning arises from the
statistical heterogeneity among clients: local datasets often come from
different but not entirely unrelated distributions, and personalization is,
therefore, necessary to achieve optimal results from each individual's
perspective. In this paper, we show how the excess risks of personalized
federated learning with a smooth, strongly convex loss depend on data
heterogeneity from a minimax point of view. Our analysis reveals a surprising
theorem of the alternative for personalized federated learning: there exists a
threshold such that (a) if a certain measure of data heterogeneity is below
this threshold, the FedAvg algorithm [McMahan et al., 2017] is minimax optimal;
(b) when the measure of heterogeneity is above this threshold, then doing pure
local training (i.e., clients solve empirical risk minimization problems on
their local datasets without any communication) is minimax optimal. As an
implication, our results show that the presumably difficult
(infinite-dimensional) problem of adapting to client-wise heterogeneity can be
reduced to a simple binary decision problem of choosing between the two
baseline algorithms. Our analysis relies on a new notion of algorithmic
stability that takes into account the nature of federated learning.
| stat |
Structure Learning of Contextual Markov Networks using Marginal
Pseudo-likelihood | Markov networks are popular models for discrete multivariate systems where
the dependence structure of the variables is specified by an undirected graph.
To allow for more expressive dependence structures, several generalizations of
Markov networks have been proposed. Here we consider the class of contextual
Markov networks which takes into account possible context-specific
independences among pairs of variables. Structure learning of contextual Markov
networks is very challenging due to the extremely large number of possible
structures. One of the main challenges has been to design a score, by which a
structure can be assessed in terms of model fit related to complexity, without
assuming chordality. Here we introduce the marginal pseudo-likelihood as an
analytically tractable criterion for general contextual Markov networks. Our
criterion is shown to yield a consistent structure estimator. Experiments
demonstrate the favorable properties of our method in terms of predictive
accuracy of the inferred models.
| stat |
Forecast Aggregation via Peer Prediction | Crowdsourcing is a popular paradigm for soliciting forecasts on future
events. As people may have different forecasts, how to aggregate solicited
forecasts into a single accurate prediction remains to be an important
challenge, especially when no historical accuracy information is available for
identifying experts. In this paper, we borrow ideas from the peer prediction
literature and assess the prediction accuracy of participants using solely the
collected forecasts. This approach leverages the correlations among peer
reports to cross-validate each participant's forecasts and allows us to assign
a "peer assessment score (PAS)" for each agent as a proxy for the agent's
prediction accuracy. We identify several empirically effective methods to
generate PAS and propose an aggregation framework that uses PAS to identify
experts and to boost existing aggregators' prediction accuracy. We evaluate our
methods over 14 real-world datasets and show that i) PAS generated from peer
prediction methods can approximately reflect the prediction accuracy of agents,
and ii) our aggregation framework demonstrates consistent and significant
improvement in the prediction accuracy over existing aggregators for both
binary and multi-choice questions under three popular accuracy measures: Brier
score (mean square error), log score (cross-entropy loss) and AUC-ROC.
| stat |
Censored Quantile Regression Forests | Random forests are powerful non-parametric regression method but are severely
limited in their usage in the presence of randomly censored observations, and
naively applied can exhibit poor predictive performance due to the incurred
biases. Based on a local adaptive representation of random forests, we develop
its regression adjustment for randomly censored regression quantile models.
Regression adjustment is based on new estimating equations that adapt to
censoring and lead to quantile score whenever the data do not exhibit
censoring. The proposed procedure named censored quantile regression forest,
allows us to estimate quantiles of time-to-event without any parametric
modeling assumption. We establish its consistency under mild model
specifications. Numerical studies showcase a clear advantage of the proposed
procedure.
| stat |
Towards a General Theory of Infinite-Width Limits of Neural Classifiers | Obtaining theoretical guarantees for neural networks training appears to be a
hard problem in a general case. Recent research has been focused on studying
this problem in the limit of infinite width and two different theories have
been developed: a mean-field (MF) and a constant kernel (NTK) limit theories.
We propose a general framework that provides a link between these seemingly
distinct theories. Our framework out of the box gives rise to a discrete-time
MF limit which was not previously explored in the literature. We prove a
convergence theorem for it and show that it provides a more reasonable
approximation for finite-width nets compared to the NTK limit if learning rates
are not very small. Also, our framework suggests a limit model that coincides
neither with the MF limit nor with the NTK one. We show that for networks with
more than two hidden layers RMSProp training has a non-trivial discrete-time MF
limit but GD training does not have one. Overall, our framework demonstrates
that both MF and NTK limits have considerable limitations in approximating
finite-sized neural nets, indicating the need for designing more accurate
infinite-width approximations for them.
| stat |
Model-based Clustering for Multivariate Networks | Network data are relational data recorded among a group of individuals, the
nodes. Multiple relations observed among the same set of nodes may be
represented by means of different networks, using a so-called multidimensional
network, or multiplex. We propose a latent space model for network data that
enables clustering of the nodes in a latent space, with clusters in this space
corresponding to communities of nodes. The clustering structure is modelled
using an infinite mixture distribution framework, which allows to perform joint
inference on the number of clusters and the cluster parameters. The method is
tested on simulated data experiments and is shown in application to a
multivariate network among students.
| stat |
New characterization based exponentiality tests for randomly censored
data | Recently, the characterization based approach for the construction of
goodness of fit tests has become popular. Most of the proposed tests have been
designed for complete i.i.d. samples. Here we present the adaptation of the
recently proposed exponentiality tests based on equidistribution-type
characterizations for the case of randomly censored data. Their asymptotic
properties are provided. Besides, we present the results of wide empirical
power study including the powers of several recent competitors. This study can
be used as a benchmark for future tests proposed for this kind of data.
| stat |
Multiple-trait Adaptive Fisher's Method for Genome-wide Association
Studies | In genome-wide association studies (GWASs), there is an increasing need for
detecting the associations between a genetic variant and multiple traits. In
studies of complex diseases, it is common to measure several potentially
correlated traits in a single GWAS. Despite the multivariate nature of the
studies, single-trait-based methods remain the most widely-adopted analysis
procedure, owing to their simplicity for studies with multiple traits as their
outcome. However, the association between a genetic variant and a single trait
sometimes can be weak, and ignoring the actual correlation among traits may
lose power. On the contrary, multiple-trait analysis, a method analyzes a group
of traits simultaneously, has been proven to be more powerful by incorporating
information from the correlated traits. Although existing methods have been
developed for multiple traits, several drawbacks limit their wide application
in GWASs. In this paper, we propose a multiple-trait adaptive Fisher's (MTAF)
method to test associations between a genetic variant and multiple traits at
once, by adaptively aggregating evidence from each trait. The proposed method
can accommodate both continuous and binary traits and it has reliable
performance under various scenarios. Using a simulation study, we compared our
proposed method with several existing methods and demonstrated its
competitiveness in terms of type I error control and statistical power. By
applying the method to the Study of Addiction: Genetics and Environment (SAGE)
dataset, we successfully identified several genes associated with substance
dependence.
| stat |
Modelling the Proliferation of Terrorism via Diffusion and Contagion | The proliferation of terrorism is a serious concern in national and
international security, as its spread is seen as an existential threat to
Western liberal democracies. Understanding and effectively modelling the spread
of terrorism provides useful insight into formulating effective responses. A
mathematical model capturing the theoretical constructs of contagion and
diffusion is constructed for explaining the spread of terrorist activity and
used to analyse data from the Global Terrorism Database from 2000--2016 for
Afghanistan, Iraq, and Israel.
| stat |
Fast Bayesian Estimation of Spatial Count Data Models | Spatial count data models are used to explain and predict the frequency of
phenomena such as traffic accidents in geographically distinct entities such as
census tracts or road segments. These models are typically estimated using
Bayesian Markov chain Monte Carlo (MCMC) simulation methods, which, however,
are computationally expensive and do not scale well to large datasets.
Variational Bayes (VB), a method from machine learning, addresses the
shortcomings of MCMC by casting Bayesian estimation as an optimisation problem
instead of a simulation problem. Considering all these advantages of VB, a VB
method is derived for posterior inference in negative binomial models with
unobserved parameter heterogeneity and spatial dependence. P\'olya-Gamma
augmentation is used to deal with the non-conjugacy of the negative binomial
likelihood and an integrated non-factorised specification of the variational
distribution is adopted to capture posterior dependencies. The benefits of the
proposed approach are demonstrated in a Monte Carlo study and an empirical
application on estimating youth pedestrian injury counts in census tracts of
New York City. The VB approach is around 45 to 50 times faster than MCMC on a
regular eight-core processor in a simulation and an empirical study, while
offering similar estimation and predictive accuracy. Conditional on the
availability of computational resources, the embarrassingly parallel
architecture of the proposed VB method can be exploited to further accelerate
its estimation by up to 20 times.
| stat |
Extended Stochastic Block Models with Application to Criminal Networks | Reliably learning group structure among nodes in network data is challenging
in modern applications. We are motivated by covert networks encoding
relationships among criminals. These data are subject to measurement errors and
exhibit a complex combination of an unknown number of core-periphery,
assortative and disassortative structures that may unveil the internal
architecture of the criminal organization. The coexistence of such noisy block
structures limits the reliability of community detection algorithms routinely
applied to criminal networks, and requires extensions of model-based solutions
to realistically characterize the node partition process, incorporate
information from node attributes, and provide improved strategies for
estimation, uncertainty quantification, model selection and prediction. To
address these goals, we develop a novel class of extended stochastic block
models (ESBM) that infer groups of nodes having common connectivity patterns
via Gibbs-type priors on the partition process. This choice encompasses several
realistic priors for criminal networks, covering solutions with fixed, random
and infinite number of possible groups, and facilitates inclusion of node
attributes in a principled manner. Among the new alternatives in our class, we
focus on the Gnedin process as a realistic prior that allows the number of
groups to be finite, random and subject to a reinforcement process coherent
with the modular structures in organized crime. A collapsed Gibbs sampler is
proposed for the whole ESBM class, and refined strategies for estimation,
prediction, uncertainty quantification and model selection are outlined. ESBM
performance is illustrated in realistic simulations and in an application to an
Italian Mafia network, where we learn key block patterns revealing a complex
hierarchical structure of the organization, mostly hidden from state-of-the-art
alternative solutions.
| stat |
A Statistical Investigation of Long Memory in Language and Music | Representation and learning of long-range dependencies is a central challenge
confronted in modern applications of machine learning to sequence data. Yet
despite the prominence of this issue, the basic problem of measuring long-range
dependence, either in a given data source or as represented in a trained deep
model, remains largely limited to heuristic tools. We contribute a statistical
framework for investigating long-range dependence in current applications of
deep sequence modeling, drawing on the well-developed theory of long memory
stochastic processes. This framework yields testable implications concerning
the relationship between long memory in real-world data and its learned
representation in a deep learning architecture, which are explored through a
semiparametric framework adapted to the high-dimensional setting.
| stat |
A Bayesian Mixture Model for Changepoint Estimation Using Ordinal
Predictors | In regression models, predictor variables with inherent ordering, such as
tumor staging ranging and ECOG performance status, are commonly seen in medical
settings. Statistically, it may be difficult to determine the functional form
of an ordinal predictor variable. Often, such a variable is dichotomized based
on whether it is above or below a certain cutoff. Other methods conveniently
treat the ordinal predictor as a continuous variable and assume a linear
relationship with the outcome. However, arbitrarily choosing a method may lead
to inaccurate inference and treatment. In this paper, we propose a Bayesian
mixture model to simultaneously assess the appropriate form of the predictor in
regression models by considering the presence of a changepoint through the lens
of a threshold detection problem. By using a mixture model framework to
consider both dichotomous and linear forms for the variable, the estimate is a
weighted average of linear and binary parameterizations. This method is
applicable to continuous, binary, and survival outcomes, and easily amenable to
penalized regression. We evaluated the proposed method using simulation studies
and apply it to two real datasets. We provide JAGS code for easy
implementation.
| stat |
Bayesian averaging of computer models with domain discrepancies: a
nuclear physics perspective | This article studies Bayesian model averaging (BMA) in the context of
competing expensive computer models in a typical nuclear physics setup. While
it is well known that BMA accounts for the additional uncertainty of the model
itself, we show that it also decreases the posterior variance of the prediction
errors via an explicit decomposition. We extend BMA to the situation where the
competing models are defined on non-identical study regions. Any model's local
forecasting difficulty is offset by predictions obtained from the average
model, thus extending individual models to the full domain. We illustrate our
methodology via pedagogical simulations and applications to forecasting nuclear
observables, which exhibit convincing improvements in both the BMA prediction
error and empirical coverage probabilities.
| stat |
Transfer Learning via $\ell_1$ Regularization | Machine learning algorithms typically require abundant data under a
stationary environment. However, environments are nonstationary in many
real-world applications. Critical issues lie in how to effectively adapt models
under an ever-changing environment. We propose a method for transferring
knowledge from a source domain to a target domain via $\ell_1$ regularization.
We incorporate $\ell_1$ regularization of differences between source parameters
and target parameters, in addition to an ordinary $\ell_1$ regularization.
Hence, our method yields sparsity for both the estimates themselves and changes
of the estimates. The proposed method has a tight estimation error bound under
a stationary environment, and the estimate remains unchanged from the source
estimate under small residuals. Moreover, the estimate is consistent with the
underlying function, even when the source estimate is mistaken due to
nonstationarity. Empirical results demonstrate that the proposed method
effectively balances stability and plasticity.
| stat |
Autocalibration and Tweedie-dominance for Insurance Pricing with Machine
Learning | Boosting techniques and neural networks are particularly effective machine
learning methods for insurance pricing. Often in practice, there are
nevertheless endless debates about the choice of the right loss function to be
used to train the machine learning model, as well as about the appropriate
metric to assess the performances of competing models. Also, the sum of fitted
values can depart from the observed totals to a large extent and this often
confuses actuarial analysts. The lack of balance inherent to training models by
minimizing deviance outside the familiar GLM with canonical link setting has
been empirically documented in W\"uthrich (2019, 2020) who attributes it to the
early stopping rule in gradient descent methods for model fitting. The present
paper aims to further study this phenomenon when learning proceeds by
minimizing Tweedie deviance. It is shown that minimizing deviance involves a
trade-off between the integral of weighted differences of lower partial moments
and the bias measured on a specific scale. Autocalibration is then proposed as
a remedy. This new method to correct for bias adds an extra local GLM step to
the analysis. Theoretically, it is shown that it implements the autocalibration
concept in pure premium calculation and ensures that balance also holds on a
local scale, not only at portfolio level as with existing bias-correction
techniques. The convex order appears to be the natural tool to compare
competing models, putting a new light on the diagnostic graphs and associated
metrics proposed by Denuit et al. (2019).
| stat |
A note on Fibonacci Sequences of Random Variables | The focus of this paper is the random sequences in the form $\{X_{0},X_{1},$
$X_{n}=X_{n-2}+X_{n-1},n=2,3,..\dot{\}},$ referred to as Fibonacci Random
Sequence (FRS). The initial random variables $X_{0}$ and $X_{1}$ are assumed to
be absolutely continuous with joint probability density function (pdf)
$f_{X_{0},X_{1}}.$ The FRS is completely determined by $X_{0}$ and $X_{1}$ and
the members of Fibonacci sequence $\digamma
\equiv\{0,1,1,2,3,5,8,13,21,34,55,89,144,...\}.$ We examine the distributional
and limit properties of the random sequence $X_{n},n=0,1,2,...$ .
| stat |
Assessment of COVID-19 hospitalization forecasts from a simplified SIR
model | We propose the SH model, a simplified version of the well-known SIR
compartmental model of infectious diseases. With optimized parameters and
initial conditions, this time-invariant two-parameter two-dimensional model is
able to fit COVID-19 hospitalization data over several months with high
accuracy (mean absolute percentage error below 15%). Moreover, we observed
that, when the model is trained on a suitable two-week period around the
hospitalization peak for Belgium, it forecasts the subsequent three-month
decrease with mean absolute percentage error below 10%. However, when it is
trained in the increase phase, it is less successful at forecasting the
subsequent evolution.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.