title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Particle rolling MCMC | An efficient simulation-based methodology is proposed for the rolling window
estimation of state space models, called particle rolling Markov chain Monte
Carlo (MCMC) with double block sampling. In our method, which is based on
Sequential Monte Carlo (SMC), particles are sequentially updated to approximate
the posterior distribution for each window by learning new information and
discarding old information from observations. Th particles are refreshed with
an MCMC algorithm when the importance weights degenerate. To avoid degeneracy,
which is crucial for reducing the computation time, we introduce a block
sampling scheme and generate multiple candidates by the algorithm based on the
conditional SMC. The theoretical discussion shows that the proposed methodology
with a nested structure is expressed as SMC sampling for the augmented space to
provide the justification. The computational performance is evaluated in
illustrative examples, showing that the posterior distributions of the model
parameters are accurately estimated. The proofs and additional discussions
(algorithms and experimental results) are provided in the Supplementary
Material.
| stat |
Risk Prediction with Imperfect Survival Outcome Information from
Electronic Health Records | Readily available proxies for time of disease onset such as time of the first
diagnostic code can lead to substantial risk prediction error if performing
analyses based on poor proxies. Due to the lack of detailed documentation and
labor intensiveness of manual annotation, it is often only feasible to
ascertain for a small subset the current status of the disease by a follow up
time rather than the exact time. In this paper, we aim to develop risk
prediction models for the onset time efficiently leveraging both a small number
of labels on current status and a large number of unlabeled observations on
imperfect proxies. Under a semiparametric transformation model for onset and a
highly flexible measurement error models for proxy onset time, we propose the
semisupervised risk prediction method by combining information from proxies and
limited labels efficiently. From an initial estimator solely based on the
labelled subset, we perform a one-step correction with the full data augmenting
against a mean zero rank correlation score derived from the proxies. We
establish the consistency and asymptotic normality of the proposed
semi-supervised estimator and provide a resampling procedure for interval
estimation. Simulation studies demonstrate that the proposed estimator performs
well in finite sample. We illustrate the proposed estimator by developing a
genetic risk prediction model for obesity using data from Partners Biobank
Electronic Health Records (EHR).
| stat |
Gonogo: An R Implementation of Test Methods to Perform, Analyze and
Simulate Sensitivity Experiments | This work provides documentation for a suite of R functions contained in
gonogo.R. The functions provide sensitivity testing practitioners and
researchers with an ability to conduct, analyze and simulate various
sensitivity experiments involving binary responses and a single stimulus level
(e.g., drug dosage, drop height, velocity, etc.). Included are the modern Neyer
and 3pod adaptive procedures, as well as the Bruceton and Langlie. The latter
two benchmark procedures are capable of being performed according to
generalized up-down transformed-response rules. Each procedure is designated
phase-one of a three-phase experiment. The goal of phase-one is to achieve
overlapping data. The two additional (and optional) refinement phases utilize
the D-optimal criteria and the Robbins-Monro-Joseph procedure. The goals of the
two refinement phases are to situate testing in the vicinity of the median and
tails of the latent response distribution, respectively.
| stat |
Auditing ML Models for Individual Bias and Unfairness | We consider the task of auditing ML models for individual bias/unfairness. We
formalize the task in an optimization problem and develop a suite of
inferential tools for the optimal value. Our tools permit us to obtain
asymptotic confidence intervals and hypothesis tests that cover the
target/control the Type I error rate exactly. To demonstrate the utility of our
tools, we use them to reveal the gender and racial biases in Northpointe's
COMPAS recidivism prediction instrument.
| stat |
Estimating the effect of PEG in ALS patients using observational data
subject to censoring by death and missing outcomes | Though they may offer valuable patient and disease information that is
impossible to study in a randomized trial, clinical disease registries also
require special care and attention in causal inference. Registry data may be
incomplete, inconsistent, and subject to confounding. In this paper we aim to
address several analytical issues in estimating treatment effects that plague
clinical registries such as the Emory amyotrophic lateral sclerosis (ALS)
Clinic Registry. When attempting to assess the effect of a surgical insertion
of a percutaneous endoscopic gastrostomy (PEG) tube on body mass index (BMI)
using the data from the ALS Clinic Registry, one must combat issues of
confounding, censoring by death, and missing outcome data that have not been
addressed in previous studies of PEG. We propose a causal inference framework
for estimating the survivor average causal effect (SACE) of PEG, which
incorporates a model for generalized propensity scores to correct for
confounding by pre-treatment variables, a model for principal stratification to
account for censoring by death, and a model for the missing data mechanism.
Applying the proposed framework to the ALS Clinic Registry Data, our analysis
shows that PEG has a positive SACE on BMI at month 18 post-baseline; our
results likely offer more definitive answers regarding the effect of PEG than
previous studies of PEG.
| stat |
Parameter Estimation with Dense and Convolutional Neural Networks
Applied to the FitzHugh-Nagumo ODE | Machine learning algorithms have been successfully used to approximate
nonlinear maps under weak assumptions on the structure and properties of the
maps. We present deep neural networks using dense and convolutional layers to
solve an inverse problem, where we seek to estimate parameters of a
FitzHugh-Nagumo model, which consists of a nonlinear system of ordinary
differential equations (ODEs). We employ the neural networks to approximate
reconstruction maps for model parameter estimation from observational data,
where the data comes from the solution of the ODE and takes the form of a time
series representing dynamically spiking membrane potential of a biological
neuron. We target this dynamical model because of the computational challenges
it poses in an inference setting, namely, having a highly nonlinear and
nonconvex data misfit term and permitting only weakly informative priors on
parameters. These challenges cause traditional optimization to fail and
alternative algorithms to exhibit large computational costs. We quantify the
prediction errors of model parameters obtained from the neural networks and
investigate the effects of network architectures with and without the presence
of noise in observational data. We generalize our framework for neural
network-based reconstruction maps to simultaneously estimate ODE parameters and
parameters of autocorrelated observational noise. Our results demonstrate that
deep neural networks have the potential to estimate parameters in dynamical
models and stochastic processes, and they are capable of predicting parameters
accurately for the FitzHugh-Nagumo model.
| stat |
Assessment of design and analysis frameworks for on-farm experimentation
through a simulation study of wheat yield in Japan | On-farm experiments can provide farmers with information on more efficient
crop management in their own fields. Developments in precision agricultural
technologies, such as yield monitoring and variable-rate application
technology, allow farmers to implement on-farm experiments. Research frameworks
including the experimental design and the statistical analysis method strongly
influences the precision of the experiment. Conventional statistical approaches
(e.g., ordinary least squares regression) may not be appropriate for on-farm
experiments because they are not capable of accurately accounting for the
underlying spatial variation in a particular response variable (e.g., yield
data). The effects of experimental designs and statistical approaches on type I
error rates and estimation accuracy were explored through a simulation study
hypothetically conducted on experiments in three wheat fields in Japan.
Isotropic and anisotropic spatial linear mixed models were established for
comparison with ordinary least squares regression models. The repeated designs
were not sufficient to reduce both the risk of a type I error and the
estimation bias on their own. A combination of a repeated design and an
anisotropic model is sometimes required to improve the precision of the
experiments. Model selection should be performed to determine whether the
anisotropic model is required for analysis of any specific field. The
anisotropic model had larger standard errors than the other models, especially
when the estimates had large biases. This finding highlights an advantage of
anisotropic models since they enable experimenters to cautiously consider the
reliability of the estimates when they have a large bias.
| stat |
Interpolation Consistency Training for Semi-Supervised Learning | We introduce Interpolation Consistency Training (ICT), a simple and
computation efficient algorithm for training Deep Neural Networks in the
semi-supervised learning paradigm. ICT encourages the prediction at an
interpolation of unlabeled points to be consistent with the interpolation of
the predictions at those points. In classification problems, ICT moves the
decision boundary to low-density regions of the data distribution. Our
experiments show that ICT achieves state-of-the-art performance when applied to
standard neural network architectures on the CIFAR-10 and SVHN benchmark
datasets. Our theoretical analysis shows that ICT corresponds to a certain type
of data-adaptive regularization with unlabeled points which reduces overfitting
to labeled points under high confidence values.
| stat |
Quantifying Performance Changes with Effect Size Confidence Intervals | Measuring performance & quantifying a performance change are core evaluation
techniques in programming language and systems research. Of 122 recent
scientific papers, as many as 65 included experimental evaluation that
quantified a performance change using a ratio of execution times. Few of these
papers evaluated their results with the level of rigour that has come to be
expected in other experimental sciences. The uncertainty of measured results
was largely ignored. Scarcely any of the papers mentioned uncertainty in the
ratio of the mean execution times, and most did not even mention uncertainty in
the two means themselves. Most of the papers failed to address the
non-deterministic execution of computer programs (caused by factors such as
memory placement, for example), and none addressed non-deterministic
compilation. It turns out that the statistical methods presented in the
computer systems performance evaluation literature for the design and summary
of experiments do not readily allow this either. This poses a hazard to the
repeatability, reproducibility and even validity of quantitative results.
Inspired by statistical methods used in other fields of science, and building
on results in statistics that did not make it to introductory textbooks, we
present a statistical model that allows us both to quantify uncertainty in the
ratio of (execution time) means and to design experiments with a rigorous
treatment of those multiple sources of non-determinism that might impact
measured performance. Better still, under our framework summaries can be as
simple as "system A is faster than system B by 5.5% $\pm$ 2.5%, with 95%
confidence", a more natural statement than those derived from typical current
practice, which are often misinterpreted.
November 2013
| stat |
Dynamic estimation with random forests for discrete-time survival data | Time-varying covariates are often available in survival studies and
estimation of the hazard function needs to be updated as new information
becomes available. In this paper, we investigate several different
easy-to-implement ways that random forests can be used for dynamic estimation
of the survival or hazard function from discrete-time survival data. The
results from a simulation study indicate that all methods can perform well, and
that none dominates the others. In general, situations that are more difficult
from an estimation point of view (such as weaker signals and less data) favour
a global fit, pooling over all time points, while situations that are easier
from an estimation point of view (such as stronger signals and more data) favor
local fits.
| stat |
Simple Regret Minimization for Contextual Bandits | There are two variants of the classical multi-armed bandit (MAB) problem that
have received considerable attention from machine learning researchers in
recent years: contextual bandits and simple regret minimization. Contextual
bandits are a sub-class of MABs where, at every time step, the learner has
access to side information that is predictive of the best arm. Simple regret
minimization assumes that the learner only incurs regret after a pure
exploration phase. In this work, we study simple regret minimization for
contextual bandits. Motivated by applications where the learner has separate
training and autonomous modes, we assume that the learner experiences a pure
exploration phase, where feedback is received after every action but no regret
is incurred, followed by a pure exploitation phase in which regret is incurred
but there is no feedback. We present the Contextual-Gap algorithm and establish
performance guarantees on the simple regret, i.e., the regret during the pure
exploitation phase. Our experiments examine a novel application to adaptive
sensor selection for magnetic field estimation in interplanetary spacecraft,
and demonstrate considerable improvement over algorithms designed to minimize
the cumulative regret.
| stat |
Scalable Gaussian Process Variational Autoencoders | Conventional variational autoencoders fail in modeling correlations between
data points due to their use of factorized priors. Amortized Gaussian process
inference through GP-VAEs has led to significant improvements in this regard,
but is still inhibited by the intrinsic complexity of exact GP inference. We
improve the scalability of these methods through principled sparse inference
approaches. We propose a new scalable GP-VAE model that outperforms existing
approaches in terms of runtime and memory footprint, is easy to implement, and
allows for joint end-to-end optimization of all components.
| stat |
Precision for binary measurement methods and results under beta-binomial
distributions | To handle typical problems from fields dealing with biological responses,
this study develops a new statistical model and method for analysing the
precision of binary measurement methods and results from collaborative studies.
The model is based on beta-binomial distributions. In other words, we assume
that the sensitivity of each laboratory obeys a beta distribution and the
binary measurement results under a given sensitivity follow a binomial
distribution. We propose the key precision indicators of repeatability and
reproducibility for the model and derive their unbiased estimates. We further
propose a confidence interval for repeatability by applying the Jeffreys
interval, which utilizes the assumption of beta distributions for sensitivity.
Moreover, we propose a statistical test for determining laboratory effects,
using simultaneous confidence intervals based on the confidence interval of
each laboratory's sensitivity. Finally, we apply the proposed method to
real-world examples in the fields of food safety and chemical risk assessment
and management.
| stat |
Multivariate Temporal Point Process Regression | Point process modeling is gaining increasing attention, as point process type
data are emerging in numerous scientific applications. In this article,
motivated by a neuronal spike trains study, we propose a novel point process
regression model, where both the response and the predictor can be a
high-dimensional point process. We model the predictor effects through the
conditional intensities using a set of basis transferring functions in a
convolutional fashion. We organize the corresponding transferring coefficients
in the form of a three-way tensor, then impose the low-rank, sparsity, and
subgroup structures on this coefficient tensor. These structures help reduce
the dimensionality, integrate information across different individual
processes, and facilitate the interpretation. We develop a highly scalable
optimization algorithm for parameter estimation. We derive the large sample
error bound for the recovered coefficient tensor, and establish the subgroup
identification consistency, while allowing the dimension of the multivariate
point process to diverge. We demonstrate the efficacy of our method through
both simulations and a cross-area neuronal spike trains analysis in a sensory
cortex study.
| stat |
Learning without feedback: Fixed random learning signals allow for
feedforward training of deep neural networks | While the backpropagation of error algorithm enables deep neural network
training, it implies (i) bidirectional synaptic weight transport and (ii)
update locking until the forward and backward passes are completed. Not only do
these constraints preclude biological plausibility, but they also hinder the
development of low-cost adaptive smart sensors at the edge, as they severely
constrain memory accesses and entail buffering overhead. In this work, we show
that the one-hot-encoded labels provided in supervised classification problems,
denoted as targets, can be viewed as a proxy for the error sign. Therefore,
their fixed random projections enable a layerwise feedforward training of the
hidden layers, thus solving the weight transport and update locking problems
while relaxing the computational and memory requirements. Based on these
observations, we propose the direct random target projection (DRTP) algorithm
and demonstrate that it provides a tradeoff between accuracy and computational
cost that is suitable for adaptive edge computing devices.
| stat |
Estimation and Feature Selection in Mixtures of Generalized Linear
Experts Models | Mixtures-of-Experts (MoE) are conditional mixture models that have shown
their performance in modeling heterogeneity in data in many statistical
learning approaches for prediction, including regression and classification, as
well as for clustering. Their estimation in high-dimensional problems is still
however challenging. We consider the problem of parameter estimation and
feature selection in MoE models with different generalized linear experts
models, and propose a regularized maximum likelihood estimation that
efficiently encourages sparse solutions for heterogeneous data with
high-dimensional predictors. The developed proximal-Newton EM algorithm
includes proximal Newton-type procedures to update the model parameter by
monotonically maximizing the objective function and allows to perform efficient
estimation and feature selection. An experimental study shows the good
performance of the algorithms in terms of recovering the actual sparse
solutions, parameter estimation, and clustering of heterogeneous regression
data, compared to the main state-of-the art competitors.
| stat |
Stylised Choropleth Maps for New Zealand Regions and District Health
Boards | New Zealand has two top-level sets of administrative divisions: the District
Health Boards and the Regions. In this note I describe a hexagonal layout for
creating stylised maps of these divisions, and using colour, size, and
triangular subdivisions to compare data between divisions and across multiple
variables. I present an implementation in the DHBins package for R using both
base graphics and ggplot2; the concepts and specific hexagonal layout could be
used in any software.
| stat |
Identifiability and estimation under the test-negative design with
population controls with the goal of identifying risk and preventive factors
for SARS-CoV-2 infection | Due to the rapidly evolving COVID-19 pandemic caused by the SARS-CoV-2 virus,
quick public health investigations of the relationships between behaviours and
infection risk are essential. Recently the test-negative design was proposed to
recruit and survey participants who are symptomatic and being tested for
SARS-CoV-2 infection with the goal of evaluating associations between the
survey responses (including behaviours and environment) and testing positive on
the test. It was also proposed to recruit additional controls who are part of
the general population as a baseline comparison group in order to evaluate risk
factors specific to SARS-CoV-2 infection. In this study, we consider an
alternative design where we recruit among all individuals, symptomatic and
asymptomatic, being tested for the virus in addition to population controls. We
define a regression parameter related to a prospective risk factor analysis and
investigate its identifiability under the two study designs. We review the
difference between the prospective risk factor parameter and the parameter
targeted in the typical test-negative design where only symptomatic and tested
people are recruited.
Using missing data directed acyclic graphs we provide conditions and required
data collection under which identifiability of the prospective risk factor
parameter is possible and compare the benefits and limitations of the
alternative study designs and target parameters. We propose a novel inverse
probability weighting estimator and demonstrate the performance of this
estimator through simulation study.
| stat |
Practical Privacy Filters and Odometers with R\'enyi Differential
Privacy and Applications to Differentially Private Deep Learning | Differential Privacy (DP) is the leading approach to privacy preserving deep
learning. As such, there are multiple efforts to provide drop-in integration of
DP into popular frameworks. These efforts, which add noise to each gradient
computation to make it DP, rely on composition theorems to bound the total
privacy loss incurred over this sequence of DP computations.
However, existing composition theorems present a tension between efficiency
and flexibility. Most theorems require all computations in the sequence to have
a predefined DP parameter, called the privacy budget. This prevents the design
of training algorithms that adapt the privacy budget on the fly, or that
terminate early to reduce the total privacy loss. Alternatively, the few
existing composition results for adaptive privacy budgets provide complex
bounds on the privacy loss, with constants too large to be practical.
In this paper, we study DP composition under adaptive privacy budgets through
the lens of R\'enyi Differential Privacy, proving a simpler composition theorem
with smaller constants, making it practical enough to use in algorithm design.
We demonstrate two applications of this theorem for DP deep learning: adapting
the noise or batch size online to improve a model's accuracy within a fixed
total privacy loss, and stopping early when fine-tuning a model to reduce total
privacy loss.
| stat |
An Exponential Efron-Stein Inequality for Lq Stable Learning Rules | There is accumulating evidence in the literature that stability of learning
algorithms is a key characteristic that permits a learning algorithm to
generalize. Despite various insightful results in this direction, there seems
to be an overlooked dichotomy in the type of stability-based generalization
bounds we have in the literature. On one hand, the literature seems to suggest
that exponential generalization bounds for the estimated risk, which are
optimal, can be only obtained through stringent, distribution independent and
computationally intractable notions of stability such as uniform stability. On
the other hand, it seems that weaker notions of stability such as hypothesis
stability, although it is distribution dependent and more amenable to
computation, can only yield polynomial generalization bounds for the estimated
risk, which are suboptimal.
In this paper, we address the gap between these two regimes of results. In
particular, the main question we address here is \emph{whether it is possible
to derive exponential generalization bounds for the estimated risk using a
notion of stability that is computationally tractable and distribution
dependent, but weaker than uniform stability. Using recent advances in
concentration inequalities, and using a notion of stability that is weaker than
uniform stability but distribution dependent and amenable to computation, we
derive an exponential tail bound for the concentration of the estimated risk of
a hypothesis returned by a general learning rule, where the estimated risk is
expressed in terms of either the resubstitution estimate (empirical error), or
the deleted (or, leave-one-out) estimate. As an illustration, we derive
exponential tail bounds for ridge regression with unbounded responses, where we
show how stability changes with the tail behavior of the response variables.
| stat |
A Generic Algorithm for Sleep-Wake Cycle Detection using Unlabeled
Actigraphy Data | One key component when analyzing actigraphy data for sleep studies is
sleep-wake cycle detection. Most detection algorithms rely on accurate sleep
diary labels to generate supervised classifiers, with parameters optimized for
a particular dataset. However, once the actigraphy trackers are deployed in the
field, labels for training models and validating detection accuracy are often
not available.
In this paper, we propose a generic, training-free algorithm to detect
sleep-wake cycles from minute-by-minute actigraphy. Leveraging a robust
nonlinear parametric model, our proposed method refines the detection region by
searching for a single change point within bounded regions defined by the
parametric model. Challenged by the absence of ground truth labels, we also
propose an evaluation metric dedicated to this problem. Tested on week-long
actigraphy from 112 children, the results show that the proposed algorithm
improves on the baseline model consistently and significantly (p<3e-15).
Moreover, focusing on the commonality in human circadian rhythm captured by
actigraphy, the proposed method is generic to data collected by various
actigraphy trackers, circumventing the laborious label collection step in
developing customized classifiers for sleep detection.
| stat |
Numerically more stable computation of the p-values for the two-sample
Kolmogorov-Smirnov test | The two-sample Kolmogorov-Smirnov test is a widely used statistical test for
detecting whether two samples are likely to come from the same distribution.
Implementations typically recur on an article of Hodges from 1957. The advances
in computation speed make it feasible to compute exact p-values for a much
larger range of problem sizes, but these run into numerical stability problems
from floating point operations. We provide a simple transformation of the
defining recurrence for the two-side two-sample KS test that avoids this.
| stat |
trialr: Bayesian Clinical Trial Designs in R and Stan | This manuscript introduces an \proglang{R} package called \pkg{trialr} that
implements a collection of clinical trial methods in \proglang{Stan} and
\proglang{R}. In this article, we explore three methods in detail. The first is
the continual reassessment method for conducting phase I dose-finding trials
that seek a maximum tolerable dose. The second is EffTox, a dose-finding design
that scrutinises doses by joint efficacy and toxicity outcomes. The third is
the augmented binary method for modelling the probability of treatment success
in phase II oncology trials with reference to repeated measures of continuous
tumour size and binary indicators of treatment failure. We emphasise in this
article the benefits that stem from having access to posterior samples,
including flexible inference and powerful visualisation. We hope that this
package encourages the use of Bayesian methods in clinical trials.
| stat |
Bi-cross validation for estimating spectral clustering hyper parameters | One challenge impeding the analysis of terabyte scale x-ray scattering data
from the Linac Coherent Light Source LCLS, is determining the number of
clusters required for the execution of traditional clustering algorithms. Here
we demonstrate that previous work using bi-cross validation (BCV) to determine
the number of singular vectors directly maps to the spectral clustering problem
of estimating both the number of clusters and hyper parameter values. These
results indicate that the process of estimating the number of clusters should
not be divorced from the process of estimating other hyper parameters. Applying
this method to LCLS x-ray scattering data enables the identification of dropped
shots without manually setting boundaries on detector fluence and provides a
path towards identifying rare and anomalous events.
| stat |
Estimating The Proportion of Signal Variables Under Arbitrary Covariance
Dependence | Estimating the proportion of signals hidden in a large amount of noise
variables is of interest in many scientific inquires. In this paper, we
consider realistic but theoretically challenging settings with arbitrary
covariance dependence between variables. We define mean absolute correlation
(MAC) to measure the overall dependence level and investigate a family of
estimators for their performances in the full range of MAC. We explicit the
joint effect of MAC dependence and signal sparsity on the performances of the
family of estimators and discover that no single estimator in the family is
most powerful under different MAC dependence levels. Informed by the
theoretical insight, we propose a new estimator to better adapt to arbitrary
covariance dependence. The proposed method compares favorably to several
existing methods in extensive finite-sample settings with strong to weak
covariance dependence and real dependence structures from genetic association
studies.
| stat |
New formulation of the Logistic-Gaussian process to analyze trajectory
tracking data | Improved communication systems, shrinking battery sizes and the price drop of
tracking devices have led to an increasing availability of trajectory tracking
data. These data are often analyzed to understand animal behavior.
In this work, we propose a new model for interpreting the animal movent as a
mixture of characteristic patterns, that we interpret as different behaviors.
The probability that the animal is behaving according to a specific pattern, at
each time instant, is non-parametrically estimated using the Logistic-Gaussian
process. Owing to a new formalization and the way we specify the
coregionalization matrix of the associated multivariate Gaussian process, our
model is invariant with respect to the choice of the reference element and of
the ordering of the probability vector components. We fit the model under a
Bayesian framework, and show that the Markov chain Monte Carlo algorithm we
propose is straightforward to implement.
We perform a simulation study with the aim of showing the ability of the
estimation procedure to retrieve the model parameters. We also test the
performance of the information criterion we used to select the number of
behaviors. The model is then applied to a real dataset where a wolf has been
observed before and after procreation. The results are easy to interpret, and
clear differences emerge in the two phases.
| stat |
Online A-Optimal Design and Active Linear Regression | We consider in this paper the problem of optimal experiment design where a
decision maker can choose which points to sample to obtain an estimate
$\hat{\beta}$ of the hidden parameter $\beta^{\star}$ of an underlying linear
model. The key challenge of this work lies in the heteroscedasticity assumption
that we make, meaning that each covariate has a different and unknown variance.
The goal of the decision maker is then to figure out on the fly the optimal way
to allocate the total budget of $T$ samples between covariates, as sampling
several times a specific one will reduce the variance of the estimated model
around it (but at the cost of a possible higher variance elsewhere). By trying
to minimize the $\ell^2$-loss $\mathbb{E}
[\lVert\hat{\beta}-\beta^{\star}\rVert^2]$ the decision maker is actually
minimizing the trace of the covariance matrix of the problem, which corresponds
then to online A-optimal design. Combining techniques from bandit and convex
optimization we propose a new active sampling algorithm and we compare it with
existing ones. We provide theoretical guarantees of this algorithm in different
settings, including a $\mathcal{O}(T^{-2})$ regret bound in the case where the
covariates form a basis of the feature space, generalizing and improving
existing results. Numerical experiments validate our theoretical findings.
| stat |
Improved inference on risk measures for univariate extremes | We discuss the use of likelihood asymptotics for inference on risk measures
in univariate extreme value problems, focusing on estimation of high quantiles
and similar summaries of risk for uncertainty quantification. We study whether
higher-order approximation based on the tangent exponential model can provide
improved inferences, and conclude that inference based on maxima is generally
robust to mild model misspecification and that profile likelihood-based
confidence intervals will often be adequate, whereas inferences based on
threshold exceedances can be badly biased but may be improved by higher-order
methods, at least for moderate sample sizes. We use the methods to shed light
on catastrophic rainfall in Venezuela, flooding in Venice, and the lifetimes of
Italian semi-supercentenarians.
| stat |
Distribution-free binary classification: prediction sets, confidence
intervals and calibration | We study three notions of uncertainty quantification -- calibration,
confidence intervals and prediction sets -- for binary classification in the
distribution-free setting, that is without making any distributional
assumptions on the data. With a focus towards calibration, we establish a
'tripod' of theorems that connect these three notions for score-based
classifiers. A direct implication is that distribution-free calibration is only
possible, even asymptotically, using a scoring function whose level sets
partition the feature space into at most countably many sets. Parametric
calibration schemes such as variants of Platt scaling do not satisfy this
requirement, while nonparametric schemes based on binning do. To close the
loop, we derive distribution-free confidence intervals for binned probabilities
for both fixed-width and uniform-mass binning. As a consequence of our 'tripod'
theorems, these confidence intervals for binned probabilities lead to
distribution-free calibration. We also derive extensions to settings with
streaming data and covariate shift.
| stat |
On Some Statistical and Axiomatic Properties of the Injury Severity
Score | The Injury Severity Score (ISS) is a standard aggregate indicator of the
overall severity of multiple injuries to the human body. This score is
calculated by summing the squares of the three highest values of the
Abbreviated Injury Scale (AIS) grades across six body regions of a trauma
victim. Despite its widespread usage over the past four decades, little is
known in the (mostly medical) literature on the subject about the axiomatic and
statistical properties of this quadratic aggregation score. To bridge this gap,
the present paper studies the ISS from the perspective of recent advances in
decision science. We demonstrate some statistical and axiomatic properties of
the ISS as a multicrtieria aggregation procedure. Our study highlights some
unintended, undesirable properties that stem from arbitrary choices in its
design and that call lead to bias in its use as a patient triage criterion.
| stat |
A Review of Dispersion Control Charts for Multivariate Individual
Observations | A multivariate control chart is designed to monitor process parameters of
multiple correlated quality characteristics. Often data on multivariate
processes are collected as individual observations, i.e. as vectors one at the
time. Various control charts have been proposed in the literature to monitor
the covariance matrix of a process when individual observations are collected.
In this study, we review this literature; we find 30 relevant articles from the
period 1987-2019. We group the articles into five categories. We observe that
less research has been done on CUSUM, high-dimensional and non-parametric type
control charts for monitoring the process covariance matrix. We describe each
proposed method, state their advantages, and limitations. Finally, we give
suggestions for future research.
| stat |
fbst: An R package for the Full Bayesian Significance Test for testing a
sharp null hypothesis against its alternative via the e-value | Hypothesis testing is a central statistical method in psychology and the
cognitive sciences. However, the problems of null hypothesis significance
testing (NHST) and p-values have been debated widely, but few attractive
alternatives exist. This article introduces the fbst R package, which
implements the Full Bayesian Significance Test (FBST) to test a sharp null
hypothesis against its alternative via the e-value. The statistical theory of
the FBST has been introduced by Pereira et al. (1999) more than two decades ago
and since then, the FBST has shown to be a Bayesian alternative to NHST and
p-values with both theoretical and practical highly appealing properties. The
algorithm provided in the fbst package is applicable to any Bayesian model as
long as the posterior distribution can be obtained at least numerically. The
core function of the package provides the Bayesian evidence against the null
hypothesis, the e-value. Additionally, p-values based on asymptotic arguments
can be computed and rich visualisations for communication and interpretation of
the results can be produced. Three examples of frequently used statistical
procedures in the cognitive sciences are given in this paper which demonstrate
how to apply the FBST in practice using the fbst package. Based on the success
of the FBST in statistical science, the fbst package should be of interest to a
broad range of researchers in psychology and the cognitive sciences and
hopefully will encourage researchers to consider the FBST as a possible
alternative when conducting hypothesis tests of a sharp null hypothesis.
| stat |
Assessing Excess Mortality in Times of Pandemics Based on Principal
Component Analysis of Weekly Mortality Data -- The Case of COVID-19 | The current outbreak of COVID-19 has called renewed attention to the need for
sound statistical analysis for monitoring mortality patterns and trends over
time. Excess mortality has been suggested as the most appropriate indicator to
measure the overall burden of the pandemic on mortality. As such, excess
mortality has received considerable interest during the first months of the
COVID-19 pandemic. Previous approaches to estimate excess mortality are
somewhat limited, as they do not include sufficiently long-term trends,
correlations among different demographic and geographic groups, and the
autocorrelations in the mortality time series. This might lead to biased
estimates of excess mortality, as random mortality fluctuations may be
misinterpreted as excess mortality. We present a blend of classical
epidemiological approaches to estimating excess mortality during extraordinary
events with an established demographic approach in mortality forecasting,
namely a Lee-Carter type model, which covers the named limitations and draws a
more realistic picture of the excess mortality. We illustrate our approach
using weekly age- and sex-specific mortality data for 19 countries and the
current COVID-19 pandemic as a case study. Our proposed model provides a
general framework that can be applied to future pandemics as well as to monitor
excess mortality from specific causes of deaths.
| stat |
Quantile Regression with Interval Data | This paper investigates the identification of quantiles and quantile
regression parameters when observations are set valued. We define the
identification set of quantiles of random sets in a way that extends the
definition of quantiles for regular random variables. We then give sharp
characterization of this set by extending concepts from random set theory. For
quantile regression parameters, we show that the identification set is
characterized by a system of conditional moment inequalities. This
characterization extends that of parametric quantile regression for regular
random variables. Estimation and inference theories are developed for
continuous cases, discrete cases, nonparametric conditional quantiles, and
parametric quantile regressions. A fast computational algorithm of set linear
programming is proposed. Monte Carlo experiments support our theoretical
properties.
| stat |
Impact of the adjustment of stratification factors on time-to-event
analyses | In a stratified clinical trial design with time to event end points,
stratification factors are often accounted for the log-rank test and the Cox
regression analyses. In this work, we have evaluated the impact of inclusion of
stratification factors on the power of the stratified log-rank test and have
compared the bias and standard error in HR estimate between multivariate and
stratified Cox regression analyses through simulation. Results from our
investigation suggests that both failing to consider stratification factor in
presence of their prognostic effect and stratification with smaller number of
events may substantially reduce the power of the log-rank test. Further, the HR
estimate from the multivariate Cox analysis is more accurate and precise
compared to the stratified Cox analysis. Our findings point towards the
necessity of evaluating the impact of stratification factors on the time to
event analyses at the time of study design which is presently not a norm.
| stat |
Relaxing door-to-door matching reduces passenger waiting times: a
workflow for the analysis of driver GPS traces in a stochastic carpooling
service | Carpooling has the potential to transform itself into a mass transportation
mode by abandoning its adherence to deterministic passenger-driver matching for
door-to-door journeys, and by adopting instead stochastic matching on a network
of fixed meeting points. Stochastic matching is where a passenger sends out a
carpooling request at a meeting point, and then waits for the arrival of a
self-selected driver who is already travelling to the requested meeting point.
Crucially there is no centrally dispatched driver. Moreover, the carpooling is
assured only between the meeting points, so the onus is on the passengers to
travel to/from them by their own means. Thus the success of a stochastic
carpooling service relies on the convergence, with minimal perturbation to
their existing travel patterns, to the meeting points which are highly
frequented by both passengers and drivers. Due to the innovative nature of
stochastic carpooling, existing off-the-shelf workflows are largely
insufficient for this purpose. To fill the gap in the market, we introduce a
novel workflow, comprising of a combination of data science and GIS (Geographic
Information Systems), to analyse driver GPS traces. We implement it for an
operational stochastic carpooling service in south-eastern France, and we
demonstrate that relaxing door-to-door matching reduces passenger waiting
times. Our workflow provides additional key operational indicators, namely the
driver flow maps, the driver flow temporal profiles and the driver
participation rates.
| stat |
Statistical and Topological Properties of Sliced Probability Divergences | The idea of slicing divergences has been proven to be successful when
comparing two probability measures in various machine learning applications
including generative modeling, and consists in computing the expected value of
a `base divergence' between one-dimensional random projections of the two
measures. However, the computational and statistical consequences of such a
technique have not yet been well-established. In this paper, we aim at bridging
this gap and derive some properties of sliced divergence functions. First, we
show that slicing preserves the metric axioms and the weak continuity of the
divergence, implying that the sliced divergence will share similar topological
properties. We then precise the results in the case where the base divergence
belongs to the class of integral probability metrics. On the other hand, we
establish that, under mild conditions, the sample complexity of the sliced
divergence does not depend on the dimension, even when the base divergence
suffers from the curse of dimensionality. We finally apply our general results
to the Wasserstein distance and Sinkhorn divergences, and illustrate our theory
on both synthetic and real data experiments.
| stat |
Seamless phase II/III clinical trials using early outcomes for treatment
or subgroup selection: Methods and aspects of their implementation | Adaptive seamless designs combine confirmatory testing, a domain of phase III
trials, with features such as treatment or subgroup selection, typically
associated with phase II trials. They promise to increase the efficiency of
development programmes of new drugs, e.g. in terms of sample size and / or
development time. It is well acknowledged that adaptive designs are more
involved from a logistical perspective and require more upfront planning, often
in form of extensive simulation studies, than conventional approaches. Here we
present a framework for adaptive treatment and subgroup selection using the
same notation, which links the somewhat disparate literature on treatment
selection on one side and on subgroup selection on the other. Furthermore, we
introduce a flexible and yet efficient simulation model that serves both
designs. As primary endpoints often take a long time to observe, interim
analyses are frequently informed by early outcomes. Therefore, all methods
presented accommodate both, interim analyses informed by the primary outcome or
an early outcome. The R package asd, previously developed to simulate designs
with treatment selection, was extended to include subpopulation selection
(so-called adaptive enrichment designs). Here we describe the functionality of
the R package asd and use it to present some worked-up examples motivated by
clinical trials in chronic obstructive pulmonary disease and oncology. The
examples illustrate various features of the R package providing at the same
time insights into the operating characteristics of adaptive seamless studies.
| stat |
Variations of Power-Expected-Posterior Priors in Normal Regression
Models | The power-expected-posterior (PEP) prior is an objective prior for Gaussian
linear models, which leads to consistent model selection inference, under the
M-closed scenario, and tends to favor parsimonious models. Recently, two new
forms of the PEP prior were proposed which generalize its applicability to a
wider range of models. The properties of these two PEP variants within the
context of the normal linear model are examined thoroughly, focusing on the
prior dispersion and on the consistency of the induced model selection
procedure. Results show that both PEP variants have larger variances than the
unit-information g-prior and that they are M-closed consistent as the limiting
behavior of the corresponding marginal likelihoods matches that of the BIC. The
consistency under the M-open case, using three different model misspecification
scenarios is further investigated.
| stat |
An Evolutionary Algorithm with Crossover and Mutation for Model-Based
Clustering | An evolutionary algorithm (EA) is developed as an alternative to the EM
algorithm for parameter estimation in model-based clustering. This EA
facilitates a different search of the fitness landscape, i.e., the likelihood
surface, utilizing both crossover and mutation. Furthermore, this EA represents
an efficient approach to "hard" model-based clustering and so it can be viewed
as a sort of generalization of the k-means algorithm, which is itself
equivalent to a restricted Gaussian mixture model. The EA is illustrated on
several datasets, and its performance is compared to other hard clustering
approaches and model-based clustering via the EM algorithm.
| stat |
Generalized Score Matching for General Domains | Estimation of density functions supported on general domains arises when the
data is naturally restricted to a proper subset of the real space. This problem
is complicated by typically intractable normalizing constants. Score matching
provides a powerful tool for estimating densities with such intractable
normalizing constants, but as originally proposed is limited to densities on
$\mathbb{R}^m$ and $\mathbb{R}_+^m$. In this paper, we offer a natural
generalization of score matching that accommodates densities supported on a
very general class of domains. We apply the framework to truncated graphical
and pairwise interaction models, and provide theoretical guarantees for the
resulting estimators. We also generalize a recently proposed method from
bounded to unbounded domains, and empirically demonstrate the advantages of our
method.
| stat |
Heckman selection-t model: parameter estimation via the EM-algorithm | Heckman selection model is perhaps the most popular econometric model in the
analysis of data with sample selection. The analyses of this model are based on
the normality assumption for the error terms, however, in some applications,
the distribution of the error term departs significantly from normality, for
instance, in the presence of heavy tails and/or atypical observation. In this
paper, we explore the Heckman selection-t model where the random errors follow
a bivariate Student's-t distribution. We develop an analytically tractable and
efficient EM-type algorithm for iteratively computing maximum likelihood
estimates of the parameters, with standard errors as a by-product. The
algorithm has closed-form expressions at the E-step, that rely on formulas for
the mean and variance of the truncated Student's-t distributions. Simulations
studies show the vulnerability of the Heckman selection-normal model, as well
as the robustness aspects of the Heckman selection-t model. Two real examples
are analyzed, illustrating the usefulness of the proposed methods. The proposed
algorithms and methods are implemented in the new R package HeckmanEM.
| stat |
Denise: Deep Learning based Robust PCA for Positive Semidefinite
Matrices | The robust PCA of high-dimensional matrices plays an essential role when
isolating key explanatory features. The currently available methods for
performing such a low-rank plus sparse decomposition are matrix specific,
meaning, the algorithm must re-run each time a new matrix should be decomposed.
Since these algorithms are computationally expensive, it is preferable to learn
and store a function that instantaneously performs this decomposition when
evaluated. Therefore, we introduce Denise, a deep learning-based algorithm for
robust PCA of symmetric positive semidefinite matrices, which learns precisely
such a function. Theoretical guarantees that Denise's architecture can
approximate the decomposition function, to arbitrary precision and with
arbitrarily high probability, are obtained. The training scheme is also shown
to convergence to a stationary point of the robust PCA's loss-function. We
train Denise on a randomly generated dataset, and evaluate the performance of
the DNN on synthetic and real-world covariance matrices. Denise achieves
comparable results to several state-of-the-art algorithms in terms of
decomposition quality, but as only one evaluation of the learned DNN is needed,
Denise outperforms all existing algorithms in terms of computation time.
| stat |
The Coordinate Sampler: A Non-Reversible Gibbs-like MCMC Sampler | In this article, we derive a novel non-reversible, continuous-time Markov
chain Monte Carlo (MCMC) sampler, called Coordinate Sampler, based on a
piecewise deterministic Markov process (PDMP), which can be seen as a variant
of the Zigzag sampler. In addition to proving a theoretical validation for this
new sampling algorithm, we show that the Markov chain it induces exhibits
geometrical ergodicity convergence, for distributions whose tails decay at
least as fast as an exponential distribution and at most as fast as a Gaussian
distribution. Several numerical examples highlight that our coordinate sampler
is more efficient than the Zigzag sampler, in terms of effective sample size.
| stat |
HINT: Hierarchical Invertible Neural Transport for Density Estimation
and Bayesian Inference | A large proportion of recent invertible neural architectures is based on a
coupling block design. It operates by dividing incoming variables into two
sub-spaces, one of which parameterizes an easily invertible (usually affine)
transformation that is applied to the other. While the Jacobian of such a
transformation is triangular, it is very sparse and thus may lack
expressiveness. This work presents a simple remedy by noting that (affine)
coupling can be repeated recursively within the resulting sub-spaces, leading
to an efficiently invertible block with dense triangular Jacobian. By
formulating our recursive coupling scheme via a hierarchical architecture, HINT
allows sampling from a joint distribution p(y,x) and the corresponding
posterior p(x|y) using a single invertible network. We demonstrate the power of
our method for density estimation and Bayesian inference on a novel data set of
2D shapes in Fourier parameterization, which enables consistent visualization
of samples for different dimensionalities.
| stat |
R = P(Y < X) for unit-Lindley distribution: inference with an
application in public health | The unit-Lindley distribution was recently introduced in the literature as a
viable alternative to the Beta and the Kumaraswamy distributions with support
in (0; 1). This distribution enjoys many virtuous properties over the named
distributions. In this article, we address the issue of parameter estimation
from a Bayesian perspective and study relative performance of different
estimators through extensive simulation studies. Significant emphasis is given
to the estimation of stress-strength reliability employing classical as well as
Bayesian approach. A non-trivial useful application in the public health domain
is presented proposing a simple metric of discrepancy.
| stat |
Improve black-box sequential anomaly detector relevancy with limited
user feedback | Anomaly detectors are often designed to catch statistical anomalies.
End-users typically do not have interest in all of the detected outliers, but
only those relevant to their application. Given an existing black-box
sequential anomaly detector, this paper proposes a method to improve its user
relevancy using a small number of human feedback. As our first contribution,
the method is agnostic to the detector: it only assumes access to its anomaly
scores, without requirement on any additional information inside it. Inspired
by a fact that anomalies are of different types, our approach identifies these
types and utilizes user feedback to assign relevancy to types. This relevancy
score, as our second contribution, is used to adjust the subsequent anomaly
selection process. Empirical results on synthetic and real-world datasets show
that our approach yields significant improvements on precision and recall over
a range of anomaly detectors.
| stat |
Search Algorithms and Loss Functions for Bayesian Clustering | We propose a randomized greedy search algorithm to find a point estimate for
a random partition based on a loss function and posterior Monte Carlo samples.
Given the large size and awkward discrete nature of the search space, the
minimization of the posterior expected loss is challenging. Our approach is a
stochastic search based on a series of greedy optimizations performed in a
random order and is embarrassingly parallel. We consider several loss
functions, including Binder loss and variation of information. We note that
criticisms of Binder loss are the result of using equal penalties of
misclassification and we show an efficient means to compute Binder loss with
potentially unequal penalties. Furthermore, we extend the original variation of
information to allow for unequal penalties and show no increased computational
costs. We provide a reference implementation of our algorithm. Using a variety
of examples, we show that our method produces clustering estimates that better
minimize the expected loss and are obtained faster than existing methods.
| stat |
The large learning rate phase of deep learning: the catapult mechanism | The choice of initial learning rate can have a profound effect on the
performance of deep networks. We present a class of neural networks with
solvable training dynamics, and confirm their predictions empirically in
practical deep learning settings. The networks exhibit sharply distinct
behaviors at small and large learning rates. The two regimes are separated by a
phase transition. In the small learning rate phase, training can be understood
using the existing theory of infinitely wide neural networks. At large learning
rates the model captures qualitatively distinct phenomena, including the
convergence of gradient descent dynamics to flatter minima. One key prediction
of our model is a narrow range of large, stable learning rates. We find good
agreement between our model's predictions and training dynamics in realistic
deep learning settings. Furthermore, we find that the optimal performance in
such settings is often found in the large learning rate phase. We believe our
results shed light on characteristics of models trained at different learning
rates. In particular, they fill a gap between existing wide neural network
theory, and the nonlinear, large learning rate, training dynamics relevant to
practice.
| stat |
A Prior for Record Linkage Based on Allelic Partitions | In database management, record linkage aims to identify multiple records that
correspond to the same individual. This task can be treated as a clustering
problem, in which a latent entity is associated with one or more noisy database
records. However, in contrast to traditional clustering applications, a large
number of clusters with a few observations per cluster is expected in this
context. In this paper, we introduce a new class of prior distributions based
on allelic partitions that is specially suited for the small cluster setting of
record linkage. Our approach makes it straightforward to introduce prior
information about the cluster size distribution at different scales, and
naturally enforces sublinear growth of the maximum cluster size -known as the
microclustering property. We also introduce a set of novel microclustering
conditions in order to impose further constraints on the cluster sizes a
priori. We evaluate the performance of our proposed class of priors using
simulated data and three official statistics data sets, and show that our
models provide competitive results compared to state-of-the-art microclustering
models in the record linkage literature. Moreover, we compare the performance
of different loss functions for optimal point estimation of the partitions
using decision-theoretical based approaches recently proposed in the
literature.
| stat |
Gradient and Newton Boosting for Classification and Regression | Boosting algorithms are frequently used in applied data science and in
research. To date, the distinction between boosting with either gradient
descent or second-order Newton updates is often not made in both applied and
methodological research, and it is thus implicitly assumed that the difference
is irrelevant. The goal of this article is to clarify this situation. In
particular, we present gradient and Newton boosting, as well as a hybrid
variant of the two, in a unified framework. We compare these boosting
algorithms with trees as base learners using various datasets and loss
functions. Our experiments show that Newton boosting outperforms gradient and
hybrid gradient-Newton boosting in terms of predictive accuracy on the majority
of datasets. We also present evidence that the reason for this is not faster
convergence of Newton boosting. In addition, we introduce a novel tuning
parameter for tree-based Newton boosting which is interpretable and important
for predictive accuracy.
| stat |
Predicting First Passage Percolation Shapes Using Neural Networks | Many random growth models have the property that the set of discovered sites,
scaled properly, converges to some deterministic set as time grows. Such
results are known as shape theorems. Typically, not much is known about the
shapes. For first passage percolation on $\mathbb{Z}^d$ we only know that the
shape is convex, compact, and inherits all the symmetries of $\mathbb{Z}^d$.
Using simulated data we construct and fit a neural network able to adequately
predict the shape of the set of discovered sites from the mean, standard
deviation, and percentiles of the distribution of the passage times. The
purpose of the note is two-fold. The main purpose is to give researchers a new
tool for \textit{quickly} getting an impression of the shape from the
distribution of the passage times -- instead of having to wait some time for
the simulations to run, as is the only available way today. The second purpose
of the note is simply to introduce modern machine learning methods into this
area of discrete probability, and a hope that it stimulates further research.
| stat |
Conditioning on the pre-test versus gain score modeling: revisiting the
controversy in a multilevel setting | We consider estimating the effect of a treatment on the progress of subjects
tested both before and after treatment assignment. A vast literature compares
the competing approaches of modeling the post-test score conditionally on the
pre-test score versus modeling the difference, namely the gain score. Our
contribution resides in analyzing the merits and drawbacks of the two
approaches in a multilevel setting. This is relevant in many fields, for
example education with students nested into schools. The multilevel structure
raises peculiar issues related to the contextual effects and the distinction
between individual-level and cluster-level treatment. We derive approximate
analytical results and compare the two approaches by a simulation study. For an
individual-level treatment our findings are in line with the literature,
whereas for a cluster-level treatment we point out the key role of the cluster
mean of the pre-test score, which favors the conditioning approach in settings
with large clusters.
| stat |
Optimal multi-wave sampling for regression modelling in two-phase
designs | Two-phase designs involve measuring extra variables on a subset of the cohort
where some variables are already measured. The goal of two-phase designs is to
choose a subsample of individuals from the cohort and analyse that subsample
efficiently. It is of interest to obtain an optimal design that gives the most
efficient estimates of regression parameters. In this paper, we propose a
multi-wave sampling design to approximate the optimal design for design-based
estimators. Influences functions are used to compute the optimal sampling
allocations. We propose to use informative priors on regression parameters to
derive the wave-1 sampling probabilities because any pre-specified sampling
probabilities may be far from optimal and decrease efficiency. Generalised
raking is used in statistical analysis. We show that a two-wave sampling with
reasonable informative priors will end up with higher precision for the
parameter of interest and be close to the underlying optimal design.
| stat |
Estimating Structural Target Functions using Machine Learning and
Influence Functions | We aim to construct a class of learning algorithms that are of practical
value to applied researchers in fields such as biostatistics, epidemiology and
econometrics, where the need to learn from incompletely observed information is
ubiquitous. We propose a new framework for statistical machine learning of
target functions arising as identifiable functionals from statistical models,
which we call `IF-learning' due to its reliance on influence functions (IFs).
This framework is problem- and model-agnostic and can be used to estimate a
broad variety of target parameters of interest in applied statistics: we can
consider any target function for which an IF of a population-averaged version
exists in analytic form. Throughout, we put particular focus on so-called
coarsening at random/doubly robust problems with partially unobserved
information. This includes problems such as treatment effect estimation and
inference in the presence of missing outcome data. Within this framework, we
propose two general learning algorithms that build on the idea of nonparametric
plug-in bias removal via IFs: the 'IF-learner' which uses pseudo-outcomes
motivated by uncentered IFs for regression in large samples and outputs entire
target functions without confidence bands, and the 'Group-IF-learner', which
outputs only approximations to a function but can give confidence estimates if
sufficient information on coarsening mechanisms is available. We apply both in
a simulation study on inferring treatment effects.
| stat |
Sparse Group Fused Lasso for Model Segmentation | This article introduces the sparse group fused lasso (SGFL) as a statistical
framework for segmenting sparse regression models with multivariate time
series. To compute solutions of the SGFL, a nonsmooth and nonseparable convex
program, we develop a hybrid optimization method that is fast, requires no
tuning parameter selection, and is guaranteed to converge to a global
minimizer. In numerical experiments, the hybrid method compares favorably to
state-of-the-art techniques with respect to computation time and numerical
accuracy; benefits are particularly substantial in high dimension. The method's
statistical performance is satisfactory in recovering nonzero regression
coefficients and excellent in change point detection. An application to air
quality data is presented. The hybrid method is implemented in the R package
sparseGFL available on the author's Github page.
| stat |
$k$-means on Positive Definite Matrices, and an Application to
Clustering in Radar Image Sequences | We state theoretical properties for $k$-means clustering of Symmetric
Positive Definite (SPD) matrices, in a non-Euclidean space, that provides a
natural and favourable representation of these data. We then provide a novel
application for this method, to time-series clustering of pixels in a sequence
of Synthetic Aperture Radar images, via their finite-lag autocovariance
matrices.
| stat |
Derandomizing Knockoffs | Model-X knockoffs is a general procedure that can leverage any feature
importance measure to produce a variable selection algorithm, which discovers
true effects while rigorously controlling the number or fraction of false
positives. Model-X knockoffs is a randomized procedure which relies on the
one-time construction of synthetic (random) variables. This paper introduces a
derandomization method by aggregating the selection results across multiple
runs of the knockoffs algorithm. The derandomization step is designed to be
flexible and can be adapted to any variable selection base procedure to yield
stable decisions without compromising statistical power. When applied to the
base procedure of Janson et al. (2016), we prove that derandomized knockoffs
controls both the per family error rate (PFER) and the k family-wise error rate
(k-FWER). Further, we carry out extensive numerical studies demonstrating tight
type-I error control and markedly enhanced power when compared with alternative
variable selection algorithms. Finally, we apply our approach to multi-stage
genome-wide association studies of prostate cancer and report locations on the
genome that are significantly associated with the disease. When
cross-referenced with other studies, we find that the reported associations
have been replicated.
| stat |
Asymptotic Performance Analysis of Non-Bayesian Quickest Change
Detection with an Energy Harvesting Sensor | In this paper, we consider a non-Bayesian sequential change detection based
on the Cumulative Sum (CUSUM) algorithm employed by an energy harvesting sensor
where the distributions before and after the change are assumed to be known. In
a slotted discrete-time model, the sensor, exclusively powered by randomly
available harvested energy, obtains a sample and computes the log-likelihood
ratio of the two distributions if it has enough energy to sense and process a
sample. If it does not have enough energy in a given slot, it waits until it
harvests enough energy to perform the task in a future time slot. We derive
asymptotic expressions for the expected detection delay (when a change actually
occurs), and the asymptotic tail distribution of the run-length to a false
alarm (when a change never happens). We show that when the average harvested
energy ($\bar H$) is greater than or equal to the energy required to sense and
process a sample ($E_s$), standard existing asymptotic results for the CUSUM
test apply since the energy storage level at the sensor is greater than $E_s$
after a sufficiently long time. However, when the $\bar H < E_s$, the energy
storage level can be modelled by a positive Harris recurrent Markov chain with
a unique stationary distribution. Using asymptotic results from Markov random
walk theory and associated nonlinear Markov renewal theory, we establish
asymptotic expressions for the expected detection delay and asymptotic
exponentiality of the tail distribution of the run-length to a false alarm in
this non-trivial case. Numerical results are provided to support the
theoretical results.
| stat |
Consistent feature selection for neural networks via Adaptive Group
Lasso | One main obstacle for the wide use of deep learning in medical and
engineering sciences is its interpretability. While neural network models are
strong tools for making predictions, they often provide little information
about which features play significant roles in influencing the prediction
accuracy. To overcome this issue, many regularization procedures for learning
with neural networks have been proposed for dropping non-significant features.
Unfortunately, the lack of theoretical results casts doubt on the applicability
of such pipelines. In this work, we propose and establish a theoretical
guarantee for the use of the adaptive group lasso for selecting important
features of neural networks. Specifically, we show that our feature selection
method is consistent for single-output feed-forward neural networks with one
hidden layer and hyperbolic tangent activation function. We demonstrate its
applicability using both simulation and data analysis.
| stat |
Rare Feature Selection in High Dimensions | It is common in modern prediction problems for many predictor variables to be
counts of rarely occurring events. This leads to design matrices in which many
columns are highly sparse. The challenge posed by such "rare features" has
received little attention despite its prevalence in diverse areas, ranging from
natural language processing (e.g., rare words) to biology (e.g., rare species).
We show, both theoretically and empirically, that not explicitly accounting for
the rareness of features can greatly reduce the effectiveness of an analysis.
We next propose a framework for aggregating rare features into denser features
in a flexible manner that creates better predictors of the response. Our
strategy leverages side information in the form of a tree that encodes feature
similarity.
We apply our method to data from TripAdvisor, in which we predict the
numerical rating of a hotel based on the text of the associated review. Our
method achieves high accuracy by making effective use of rare words; by
contrast, the lasso is unable to identify highly predictive words if they are
too rare. A companion R package, called rare, implements our new estimator,
using the alternating direction method of multipliers.
| stat |
Training Two-Layer ReLU Networks with Gradient Descent is Inconsistent | We prove that two-layer (Leaky)ReLU networks initialized by e.g. the widely
used method proposed by He et al. (2015) and trained using gradient descent on
a least-squares loss are not universally consistent. Specifically, we describe
a large class of one-dimensional data-generating distributions for which, with
high probability, gradient descent only finds a bad local minimum of the
optimization landscape. It turns out that in these cases, the found network
essentially performs linear regression even if the target function is
non-linear. We further provide numerical evidence that this happens in
practical situations, for some multi-dimensional distributions and that
stochastic gradient descent exhibits similar behavior.
| stat |
Ranking, and other Properties, of Elite Swimmers using Extreme Value
Theory | The International Swimming Federation (FINA) uses a very simple points system
with the aim to rank swimmers across all Olympic events. The points acquired is
a function of the ratio of the recorded time and the current world record for
that event. With some world records considered "better" than others however,
bias is introduced between events, with some being much harder to attain points
where the world record is hard to beat. A model based on extreme value theory
will be introduced, where swim-times are modelled through their rate of
occurrence, and with the distribution of the best times following a generalised
Pareto distribution. Within this framework, the strength of a particular swim
is judged based on its position compared to the whole distribution of
swim-times, rather than just the world record. This model also accounts for the
date of the swim, as training methods improve over the years, as well as
changes in technology, such as full body suits. The parameters of the
generalised Pareto distribution, for each of the 34 individual Olympic events,
will be shown to vary with a covariate, leading to a novel single unified
description of swim quality over all events and time. This structure, which
allows information to be shared across all strokes, distances, and genders,
improves the predictive power as well as the model robustness compared to
equivalent independent models. A by-product of the model is that it is possible
to estimate other features of interest, such as the ultimate possible time, the
distribution of new world records for any event, and to correct swim times for
the effect of full body suits. The methods will be illustrated using a dataset
of the best 500 swim-times for each event in the period 2001-2018.
| stat |
A Contrastive Divergence for Combining Variational Inference and MCMC | We develop a method to combine Markov chain Monte Carlo (MCMC) and
variational inference (VI), leveraging the advantages of both inference
approaches. Specifically, we improve the variational distribution by running a
few MCMC steps. To make inference tractable, we introduce the variational
contrastive divergence (VCD), a new divergence that replaces the standard
Kullback-Leibler (KL) divergence used in VI. The VCD captures a notion of
discrepancy between the initial variational distribution and its improved
version (obtained after running the MCMC steps), and it converges
asymptotically to the symmetrized KL divergence between the variational
distribution and the posterior of interest. The VCD objective can be optimized
efficiently with respect to the variational parameters via stochastic
optimization. We show experimentally that optimizing the VCD leads to better
predictive performance on two latent variable models: logistic matrix
factorization and variational autoencoders (VAEs).
| stat |
Optimal Bayesian design for model discrimination via classification | Performing optimal Bayesian design for discriminating between competing
models is computationally intensive as it involves estimating posterior model
probabilities for thousands of simulated datasets. This issue is compounded
further when the likelihood functions for the rival models are computationally
expensive. A new approach using supervised classification methods is developed
to perform Bayesian optimal model discrimination design. This approach requires
considerably fewer simulations from the candidate models than previous
approaches using approximate Bayesian computation. Further, it is easy to
assess the performance of the optimal design through the misclassification
error rate. The approach is particularly useful in the presence of models with
intractable likelihoods but can also provide computational advantages when the
likelihoods are manageable.
| stat |
Factor copula models for mixed data | We develop factor copula models for analysing the dependence among mixed
continuous and discrete responses. Factor copula models are canonical vine
copulas that involve both observed and latent variables, hence they allow tail,
asymmetric and non-linear dependence. They can be explained as conditional
independence models with latent variables that don't necessarily have an
additive latent structure. We focus on important issues that would interest the
social data analyst, such as model selection and goodness-of-fit. Our general
methodology is demonstrated with an extensive simulation study and illustrated
by re-analysing three mixed response datasets. Our study suggests that there
can be a substantial improvement over the standard factor model for mixed data
and makes the argument for moving to factor copula models.
| stat |
GPflux: A Library for Deep Gaussian Processes | We introduce GPflux, a Python library for Bayesian deep learning with a
strong emphasis on deep Gaussian processes (DGPs). Implementing DGPs is a
challenging endeavour due to the various mathematical subtleties that arise
when dealing with multivariate Gaussian distributions and the complex
bookkeeping of indices. To date, there are no actively maintained, open-sourced
and extendable libraries available that support research activities in this
area. GPflux aims to fill this gap by providing a library with state-of-the-art
DGP algorithms, as well as building blocks for implementing novel Bayesian and
GP-based hierarchical models and inference schemes. GPflux is compatible with
and built on top of the Keras deep learning eco-system. This enables
practitioners to leverage tools from the deep learning community for building
and training customised Bayesian models, and create hierarchical models that
consist of Bayesian and standard neural network layers in a single coherent
framework. GPflux relies on GPflow for most of its GP objects and operations,
which makes it an efficient, modular and extensible library, while having a
lean codebase.
| stat |
Modelling Air Pollution Crises Using Multi-agent Simulation | This paper describes an agent based approach for simulating the control of an
air pollution crisis. A Gaussian Plum air pollution dispersion model (GPD) is
combined with an Artificial Neural Network (ANN) to predict the concentration
levels of three different air pollutants. The two models (GPM and ANN) are
integrated with a MAS (multi-agent system). The MAS models pollutant sources
controllers and air pollution monitoring agencies as software agents. The
population of agents cooperates with each other in order to reduce their
emissions and control the air pollution. Leaks or natural sources of pollution
are modelled as uncontrolled sources. A cooperation strategy is simulated and
its impact on air pollution evolution is assessed and compared. The simulation
scenario is built using data about Annaba (a city in NorthEast Algeria). The
simulation helps to compare and assess the efficiency of policies to control
air pollution during crises, and takes in to account uncontrolled sources.
| stat |
Relations Between Adjacency and Modularity Graph Partitioning | In this paper the exact linear relation between the leading eigenvector of
the unnormalized modularity matrix and the eigenvectors of the adjacency matrix
is developed. Based on this analysis a method to approximate the leading
eigenvector of the modularity matrix is given, and the relative error of the
approximation is derived. A complete proof of the equivalence between
normalized modularity clustering and normalized adjacency clustering is also
given. Some applications and experiments are given to illustrate and
corroborate the points that are made in the theoretical development.
| stat |
Variable selection in sparse GLARMA models | In this paper, we propose a novel and efficient two-stage variable selection
approach for sparse GLARMA models, which are pervasive for modeling
discrete-valued time series. Our approach consists in iteratively combining the
estimation of the autoregressive moving average (ARMA) coefficients of GLARMA
models with regularized methods designed for performing variable selection in
regression coefficients of Generalized Linear Models (GLM). We first establish
the consistency of the ARMA part coefficient estimators in a specific case.
Then, we explain how to efficiently implement our approach. Finally, we assess
the performance of our methodology using synthetic data and compare it with
alternative methods. Our approach is very attractive since it benefits from a
low computational load and is able to outperform the other methods in terms of
coefficient estimation, particularly in recovering the non null regression
coefficients.
| stat |
Do In-Person Lectures Help? A Study of a Large Statistics Class | Over 1000 students over four semesters were given the option of taking an
introductory statistics class either by in-person attendance in lectures,
augmented by online recorded lectures, or by taking the same class without the
in-person lectures. Roughly equal numbers of students chose each option. The
all-online students did slightly better on computer-graded exams. The causal
effect of choosing only online lectures was estimated by adjusting for
potential confounders, most importantly the incoming ACT math scores, using
four methods. The four nearly identical point estimates remained positive but
were very small and not statistically significant. No statistically significant
differences were found in preliminary comparisons of effects on females/males,
U.S./non-U.S. citizens, freshmen/non-freshman, and lower-scoring/higher-scoring
math ACT groups.
| stat |
Selecting the Derivative of a Functional Covariate in Scalar-on-Function
Regression | This paper presents tests to formally choose between regression models using
different derivatives of a functional covariate in scalar-on-function
regression. We demonstrate that for linear regression, models using different
derivatives can be nested within a model that includes point-impact effects at
the end-points of the observed functions. Contrasts can then be employed to
test the specification of different derivatives. When nonlinear regression
models are defined, we apply a $J$ test to determine the statistical
significance of the nonlinear structure between a functional covariate and a
scalar response. The finite-sample performance of these methods is verified in
simulation, and their practical application is demonstrated using a chemometric
data set.
| stat |
Supervised variable selection in randomised controlled trials prior to
exploration of treatment effect heterogeneity: an example from severe malaria | Exploration of treatment effect heterogeneity (TEH) is an increasingly
important aspect of modern statistical analysis for stratified medicine in
randomised controlled trials (RCTs) as we start to gather more information on
trial participants and wish to maximise the opportunities for learning from
data. However, the analyst should refrain from including a large number of
variables in a treatment interaction discovery stage. Because doing so can
significantly dilute the power to detect any true outcome-predictive
interactions between treatments and covariates. Current guidance is limited and
mainly relies on the use of unsupervised learning methods, such as hierarchical
clustering or principal components analysis, to reduce the dimension of the
variable space prior to interaction tests. In this article we show that
outcome-driven dimension reduction, i.e. supervised variable selection, can
maintain power without inflating the type-I error or false-positive rate. We
provide theoretical and applied results to support our approach. The applied
results are obtained from illustrating our framework on the dataset from an RCT
in severe malaria. We also pay particular attention to the internal risk model
approach for TEH discovery, which we show is a particular case of our method
and we point to improvements over current implementation.
| stat |
An empirical comparison and characterisation of nine popular clustering
methods | Nine popular clustering methods are applied to 42 real data sets. The aim is
to give a detailed characterisation of the methods by means of several cluster
validation indexes that measure various individual aspects of the resulting
clusters such as small within-cluster distances, separation of clusters,
closeness to a Gaussian distribution etc. as introduced in Hennig (2019). 30 of
the data sets come with a "true" clustering. On these data sets the similarity
of the clusterings from the nine methods to the "true" clusterings is explored.
Furthermore, a mixed effects regression relates the observable individual
aspects of the clusters to the similarity with the "true" clusterings, which in
real clustering problems is unobservable. The study gives new insight not only
into the ability of the methods to discover "true" clusterings, but also into
properties of clusterings that can be expected from the methods, which is
crucial for the choice of a method in a real situation without a given "true"
clustering.
| stat |
Goodness-of-fit tests for functional linear models based on integrated
projections | Functional linear models are one of the most fundamental tools to assess the
relation between two random variables of a functional or scalar nature. This
contribution proposes a goodness-of-fit test for the functional linear model
with functional response that neatly adapts to functional/scalar
responses/predictors. In particular, the new goodness-of-fit test extends a
previous proposal for scalar response. The test statistic is based on a
convenient regularized estimator, is easy to compute, and is calibrated through
an efficient bootstrap resampling. A graphical diagnostic tool, useful to
visualize the deviations from the model, is introduced and illustrated with a
novel data application. The R package goffda implements the proposed methods
and allows for the reproducibility of the data application.
| stat |
Correlating Paleoclimate Time Series: Sources of Uncertainty and
Potential Pitfalls | Comparing paleoclimate time series is complicated by a variety of typical
features, including irregular sampling, age model uncertainty (e.g., errors due
to interpolation between radiocarbon sampling points) and time uncertainty
(uncertainty in calibration), which, taken together, result in unequal and
uncertain observation times of the individual time series to be correlated.
Several methods have been proposed to approximate the joint probability
distribution needed to estimate correlations, most of which rely either on
interpolation or temporal downsampling.
Here, we compare the performance of some popular approximation methods using
synthetic data resembling common properties of real world marine sediment
records. Correlations are determined by estimating the parameters of a
bivariate Gaussian model from the data using Markov Chain Monte Carlo sampling.
We complement our pseudoproxy experiments by applying the same methodology to a
pair of marine benthic oxygen records from the Atlantic Ocean.
We find that methods based upon interpolation yield better results in terms
of precision and accuracy than those which reduce the number of observations.
In all cases, the specific characteristics of the studied time series are,
however, more important than the choice of a particular interpolation method.
Relevant features include the number of observations, the persistence of each
record, and the imposed coupling strength between the paired series. In most of
our pseudoproxy experiments, uncertainty in observation times introduces less
additional uncertainty than unequal sampling and errors in observation times
do. Thus, it can be reasonable to rely on published time scales as long as
calibration uncertainties are not known.
| stat |
Testing Rank of Incomplete Unimodal Matrices | Several statistics-based detectors, based on unimodal matrix models, for
determining the number of sources in a field are designed. A new variance ratio
statistic is proposed, and its asymptotic distribution is analyzed. The
variance ratio detector is shown to outperform the alternatives. It is shown
that further improvements are achievable via optimally selected rotations.
Numerical experiments demonstrate the performance gains of our detection
methods over the baseline approach.
| stat |
cissa(): A MATLAB Function for Signal Extraction | cissa() is a MATLAB function for signal extraction by Circulant Singular
Spectrum Analysis, a procedure proposed in Bogalo et al (2021). cissa()
extracts the underlying signals in a time series identifying their frequency of
oscillation in an automated way, by just introducing the data and the window
length. This solution can be applied to stationary as well as to non-stationary
and non-linear time series. Additionally, in this paper, we solve some
technical issues regarding the beginning and end of sample data points. We also
introduce novel criteria in order to reconstruct the underlying signals
grouping some of the extracted components. The output of cissa() is the input
of the function group() to reconstruct the desired signals by further grouping
the extracted components. group() allows a novel user to create standard
signals by automated grouping options while an expert user can decide on the
number of groups and their composition. To illustrate its versatility and
performance in several fields we include 3 examples: an AM-FM synthetic signal,
an example of the physical world given by a voiced speech signal and an
economic time series. Possible applications include de-noising,
de-seasonalizing, de-trending and extracting business cycles, among others.
| stat |
Incorporating baseline covariates to validate surrogate endpoints with a
constant biomarker under control arm | A surrogate endpoint S in a clinical trial is an outcome that may be measured
earlier or more easily than the true outcome of interest T. In this work, we
extend causal inference approaches to validate such a surrogate using potential
outcomes. The causal association paradigm assesses the relationship of the
treatment effect on the surrogate with the treatment effect on the true
endpoint. Using the principal surrogacy criteria, we utilize the joint
conditional distribution of the potential outcomes T, given the potential
outcomes S. In particular, our setting of interest allows us to assume the
surrogate under the placebo, S(0), is zero-valued, and we incorporate baseline
covariates in the setting of normally-distributed endpoints. We develop
Bayesian methods to incorporate conditional independence and other modeling
assumptions and explore their impact on the assessment of surrogacy. We
demonstrate our approach via simulation and data that mimics an ongoing study
of a muscular dystrophy gene therapy.
| stat |
Physics Informed Deep Learning for Transport in Porous Media. Buckley
Leverett Problem | We present a new hybrid physics-based machine-learning approach to reservoir
modeling. The methodology relies on a series of deep adversarial neural network
architecture with physics-based regularization. The network is used to simulate
the dynamic behavior of physical quantities (i.e. saturation) subject to a set
of governing laws (e.g. mass conservation) and corresponding boundary and
initial conditions. A residual equation is formed from the governing
partial-differential equation and used as part of the training. Derivatives of
the estimated physical quantities are computed using automatic differentiation
algorithms. This allows the model to avoid overfitting, by reducing the
variance and permits extrapolation beyond the range of the training data
including uncertainty implicitely derived from the distribution output of the
generative adversarial networks. The approach is used to simulate a 2 phase
immiscible transport problem (Buckley Leverett). From a very limited dataset,
the model learns the parameters of the governing equation and is able to
provide an accurate physical solution, both in terms of shock and rarefaction.
We demonstrate how this method can be applied in the context of a forward
simulation for continuous problems. The use of these models for the inverse
problem is also presented, where the model simultaneously learns the physical
laws and determines key uncertainty subsurface parameters. The proposed
methodology is a simple and elegant way to instill physical knowledge to
machine-learning algorithms. This alleviates the two most significant
shortcomings of machine-learning algorithms: the requirement for large datasets
and the reliability of extrapolation. The principles presented in this paper
can be generalized in innumerable ways in the future and should lead to a new
class of algorithms to solve both forward and inverse physical problems.
| stat |
Inter-domain Deep Gaussian Processes | Inter-domain Gaussian processes (GPs) allow for high flexibility and low
computational cost when performing approximate inference in GP models. They are
particularly suitable for modeling data exhibiting global structure but are
limited to stationary covariance functions and thus fail to model
non-stationary data effectively. We propose Inter-domain Deep Gaussian
Processes, an extension of inter-domain shallow GPs that combines the
advantages of inter-domain and deep Gaussian processes (DGPs), and demonstrate
how to leverage existing approximate inference methods to perform simple and
scalable approximate inference using inter-domain features in DGPs. We assess
the performance of our method on a range of regression tasks and demonstrate
that it outperforms inter-domain shallow GPs and conventional DGPs on
challenging large-scale real-world datasets exhibiting both global structure as
well as a high-degree of non-stationarity.
| stat |
Joint Tracking of Multiple Quantiles Through Conditional Quantiles | Estimation of quantiles is one of the most fundamental real-time analysis
tasks. Most real-time data streams vary dynamically with time and incremental
quantile estimators document state-of-the art performance to track quantiles of
such data streams. However, most are not able to make joint estimates of
multiple quantiles in a consistent manner, and estimates may violate the
monotone property of quantiles. In this paper we propose the general concept of
*conditional quantiles* that can extend incremental estimators to jointly track
multiple quantiles. We apply the concept to propose two new estimators.
Extensive experimental results, on both synthetic and real-life data, show that
the new estimators clearly outperform legacy state-of-the-art joint quantile
tracking algorithm and achieve faster adaptivity in dynamically varying data
streams.
| stat |
A General Framework for Survival Analysis and Multi-State Modelling | Survival models are a popular tool for the analysis of time to event data
with applications in medicine, engineering, economics, and many more. Advances
like the Cox proportional hazard model have enabled researchers to better
describe hazard rates for the occurrence of single fatal events, but are unable
to accurately model competing events and transitions. Common phenomena are
often better described through multiple states, for example: the progress of a
disease modeled as healthy, sick and dead instead of healthy and dead, where
the competing nature of death and disease has to be taken into account.
Moreover, Cox models are limited by modeling assumptions, like proportionality
of hazard rates and linear effects. Individual characteristics can vary
significantly between observational units, like patients, resulting in
idiosyncratic hazard rates and different disease trajectories. These
considerations require flexible modeling assumptions. To overcome these issues,
we propose the use of neural ordinary differential equations as a flexible and
general method for estimating multi-state survival models by directly solving
the Kolmogorov forward equations. To quantify the uncertainty in the resulting
individual cause-specific hazard rates, we further introduce a variational
latent variable model and show that this enables meaningful clustering with
respect to multi-state outcomes as well as interpretability regarding covariate
values. We show that our model exhibits state-of-the-art performance on popular
survival data sets and demonstrate its efficacy in a multi-state setting
| stat |
Stratified stochastic variational inference for high-dimensional network
factor model | There has been considerable recent interest in Bayesian modeling of
high-dimensional networks via latent space approaches. When the number of nodes
increases, estimation based on Markov Chain Monte Carlo can be extremely slow
and show poor mixing, thereby motivating research on alternative algorithms
that scale well in high-dimensional settings. In this article, we focus on the
latent factor model, a widely used approach for latent space modeling of
network data. We develop scalable algorithms to conduct approximate Bayesian
inference via stochastic optimization. Leveraging sparse representations of
network data, the proposed algorithms show massive computational and storage
benefits, and allow to conduct inference in settings with thousands of nodes.
| stat |
Malaria Incidence in the Philippines: Prediction using the
Autoregressive Moving Average Models | The study was conducted to develop an appropriate model that could predict
the weekly reported Malaria incidence in the Philippines using the Box-Jenkins
method.The data were retrieved from the Department of Health(DOH) website in
the Philippines. It contains 70 data points of which 60 data points were used
in model building and the remaining 10 data points were used for forecast
evaluation. The R Statistical Software was used to do all the necessary
computations in the study. Box-Cox Transformation and Differencing was done to
make the series stationary. Based on the results of the analysis, ARIMA (2, 1,
0) is the appropriate model for the weekly Malaria incidence in the
Philippines.
| stat |
Overcomplete Independent Component Analysis via SDP | We present a novel algorithm for overcomplete independent components analysis
(ICA), where the number of latent sources k exceeds the dimension p of observed
variables. Previous algorithms either suffer from high computational complexity
or make strong assumptions about the form of the mixing matrix. Our algorithm
does not make any sparsity assumption yet enjoys favorable computational and
theoretical properties. Our algorithm consists of two main steps: (a)
estimation of the Hessians of the cumulant generating function (as opposed to
the fourth and higher order cumulants used by most algorithms) and (b) a novel
semi-definite programming (SDP) relaxation for recovering a mixing component.
We show that this relaxation can be efficiently solved with a projected
accelerated gradient descent method, which makes the whole algorithm
computationally practical. Moreover, we conjecture that the proposed program
recovers a mixing component at the rate k < p^2/4 and prove that a mixing
component can be recovered with high probability when k < (2 - epsilon) p log p
when the original components are sampled uniformly at random on the hyper
sphere. Experiments are provided on synthetic data and the CIFAR-10 dataset of
real images.
| stat |
Understanding MCMC Dynamics as Flows on the Wasserstein Space | It is known that the Langevin dynamics used in MCMC is the gradient flow of
the KL divergence on the Wasserstein space, which helps convergence analysis
and inspires recent particle-based variational inference methods (ParVIs). But
no more MCMC dynamics is understood in this way. In this work, by developing
novel concepts, we propose a theoretical framework that recognizes a general
MCMC dynamics as the fiber-gradient Hamiltonian flow on the Wasserstein space
of a fiber-Riemannian Poisson manifold. The "conservation + convergence"
structure of the flow gives a clear picture on the behavior of general MCMC
dynamics. The framework also enables ParVI simulation of MCMC dynamics, which
enriches the ParVI family with more efficient dynamics, and also adapts ParVI
advantages to MCMCs. We develop two ParVI methods for a particular MCMC
dynamics and demonstrate the benefits in experiments.
| stat |
Exploring Consequences of Simulation Design for Apparent Performance of
Statistical Methods. 1: Results from simulations with constant sample sizes | Contemporary statistical publications rely on simulation to evaluate
performance of new methods and compare them with established methods. In the
context of meta-analysis of log-odds-ratios, we investigate how the ways in
which simulations are implemented affect such conclusions. Choices of
distributions for sample sizes and/or control probabilities considerably affect
conclusions about statistical methods. Here we report on the results for
constant sample sizes. Our two subsequent publications will cover normally and
uniformly distributed sample sizes.
| stat |
Assessing the Significance of Directed and Multivariate Measures of
Linear Dependence Between Time Series | Inferring linear dependence between time series is central to our
understanding of natural and artificial systems. Unfortunately, the hypothesis
tests that are used to determine statistically significant directed or
multivariate relationships from time-series data often yield spurious
associations (Type I errors) or omit causal relationships (Type II errors).
This is due to the autocorrelation present in the analysed time series -- a
property that is ubiquitous across diverse applications, from brain dynamics to
climate change. Here we show that, for limited data, this issue cannot be
mediated by fitting a time-series model alone (e.g., in Granger causality or
prewhitening approaches), and instead that the degrees of freedom in
statistical tests should be altered to account for the effective sample size
induced by cross-correlations in the observations. This insight enabled us to
derive modified hypothesis tests for any multivariate correlation-based
measures of linear dependence between covariance-stationary time series,
including Granger causality and mutual information with Gaussian marginals. We
use both numerical simulations (generated by autoregressive models and digital
filtering) as well as recorded fMRI-neuroimaging data to show that our tests
are unbiased for a variety of stationary time series. Our experiments
demonstrate that the commonly used $F$- and $\chi^2$-tests can induce
significant false-positive rates of up to $100\%$ for both measures, with and
without prewhitening of the signals. These findings suggest that many
dependencies reported in the scientific literature may have been, and may
continue to be, spuriously reported or missed if modified hypothesis tests are
not used when analysing time series.
| stat |
brolgar: An R package to BRowse Over Longitudinal Data Graphically and
Analytically in R | Longitudinal (panel) data provide the opportunity to examine temporal
patterns of individuals, because measurements are collected on the same person
at different, and often irregular, time points. The data is typically
visualised using a "spaghetti plot", where a line plot is drawn for each
individual. When overlaid in one plot, it can have the appearance of a bowl of
spaghetti. With even a small number of subjects, these plots are too overloaded
to be read easily. The interesting aspects of individual differences are lost
in the noise. Longitudinal data is often modelled with a hierarchical linear
model to capture the overall trends, and variation among individuals, while
accounting for various levels of dependence. However, these models can be
difficult to fit, and can miss unusual individual patterns. Better visual tools
can help to diagnose longitudinal models, and better capture the individual
experiences. This paper introduces the R package, brolgar (BRowse over
Longitudinal data Graphically and Analytically in R), which provides tools to
identify and summarise interesting individual patterns in longitudinal data.
| stat |
Linear Regression with Distributed Learning: A Generalization Error
Perspective | Distributed learning provides an attractive framework for scaling the
learning task by sharing the computational load over multiple nodes in a
network. Here, we investigate the performance of distributed learning for
large-scale linear regression where the model parameters, i.e., the unknowns,
are distributed over the network. We adopt a statistical learning approach. In
contrast to works that focus on the performance on the training data, we focus
on the generalization error, i.e., the performance on unseen data. We provide
high-probability bounds on the generalization error for both isotropic and
correlated Gaussian data as well as sub-gaussian data. These results reveal the
dependence of the generalization performance on the partitioning of the model
over the network. In particular, our results show that the generalization error
of the distributed solution can be substantially higher than that of the
centralized solution even when the error on the training data is at the same
level for both the centralized and distributed approaches. Our numerical
results illustrate the performance with both real-world image data as well as
synthetic data.
| stat |
A Multiclass Classification Approach to Label Ranking | In multiclass classification, the goal is to learn how to predict a random
label $Y$, valued in $\mathcal{Y}=\{1,\; \ldots,\; K \}$ with $K\geq 3$, based
upon observing a r.v. $X$, taking its values in $\mathbb{R}^q$ with $q\geq 1$
say, by means of a classification rule $g:\mathbb{R}^q\to \mathcal{Y}$ with
minimum probability of error $\mathbb{P}\{Y\neq g(X) \}$. However, in a wide
variety of situations, the task targeted may be more ambitious, consisting in
sorting all the possible label values $y$ that may be assigned to $X$ by
decreasing order of the posterior probability $\eta_y(X)=\mathbb{P}\{Y=y \mid X
\}$. This article is devoted to the analysis of this statistical learning
problem, halfway between multiclass classification and posterior probability
estimation (regression) and referred to as label ranking here. We highlight the
fact that it can be viewed as a specific variant of ranking median regression
(RMR), where, rather than observing a random permutation $\Sigma$ assigned to
the input vector $X$ and drawn from a Bradley-Terry-Luce-Plackett model with
conditional preference vector $(\eta_1(X),\; \ldots,\; \eta_K(X))$, the sole
information available for training a label ranking rule is the label $Y$ ranked
on top, namely $\Sigma^{-1}(1)$. Inspired by recent results in RMR, we prove
that under appropriate noise conditions, the One-Versus-One (OVO) approach to
multiclassification yields, as a by-product, an optimal ranking of the labels
with overwhelming probability. Beyond theoretical guarantees, the relevance of
the approach to label ranking promoted in this article is supported by
experimental results.
| stat |
Fixing Match-Fixing: Optimal schedules to promote competitiveness | In the last round of the FIFA World Cup group stage, games for which the
outcome does not affect the selection of the qualified teams are played with
little enthusiasm. Furthermore, a team that has already qualified may take into
account other factors, such as the opponents it will face in the next stage of
the competition so that, depending on the results in the other groups and the
scheduling of the next stage, winning the game may not be in its best interest.
Even more critically, there may be situations in which a simple draw will
qualify both teams for the next stage of the competition. Any situation in
which the two opposing teams do not play competitively is detrimental to the
sport, and, above all, can lead to collusion and match-fixing opportunities. We
here develop a relatively general method of evaluating competitiveness and
apply it to the current format of the World Cup group stage. We then propose
changes to the current format in order to increase the stakes in the last round
of games of the group stage, making games more exciting to watch and, at the
same time, reducing any collusion opportunities. We appeal to the same method
to evaluate a "groups of 3" format which will be introduced in the 2026 World
Cup edition as well as a format similar to the one of the current Euro UEFA
Cup.
| stat |
A decision integration strategy for short-term demand forecasting and
ordering for red blood cell components | Blood transfusion is one of the most crucial and commonly administered
therapeutics worldwide. The need for more accurate and efficient ways to manage
blood demand and supply is an increasing concern. Building a technology-based,
robust blood demand and supply chain that can achieve the goals of reducing
ordering frequency, inventory level, wastage and shortage, while maintaining
the safety of blood usage, is essential in modern healthcare systems. In this
study, we summarize the key challenges in current demand and supply management
for red blood cells (RBCs). We combine ideas from statistical time series
modeling, machine learning, and operations research in developing an ordering
decision strategy for RBCs, through integrating a hybrid demand forecasting
model using clinical predictors and a data-driven multi-period inventory
problem considering inventory and reorder constraints. We have applied the
integrated ordering strategy to the blood inventory management system in
Hamilton, Ontario using a large clinical database from 2008 to 2018. The
proposed hybrid demand forecasting model provides robust and accurate
predictions, and identifies important clinical predictors for short-term RBC
demand forecasting. Compared with the actual historical data, our integrated
ordering strategy reduces the inventory level by 40% and decreases the ordering
frequency by 60%, with low incidence of shortages and wastage due to
expiration. If implemented successfully, our proposed strategy can achieve
significant cost savings for healthcare systems and blood suppliers. The
proposed ordering strategy is generalizable to other blood products or even
other perishable products.
| stat |
Transferable Clean-Label Poisoning Attacks on Deep Neural Nets | Clean-label poisoning attacks inject innocuous looking (and "correctly"
labeled) poison images into training data, causing a model to misclassify a
targeted image after being trained on this data. We consider transferable
poisoning attacks that succeed without access to the victim network's outputs,
architecture, or (in some cases) training data. To achieve this, we propose a
new "polytope attack" in which poison images are designed to surround the
targeted image in feature space. We also demonstrate that using Dropout during
poison creation helps to enhance transferability of this attack. We achieve
transferable attack success rates of over 50% while poisoning only 1% of the
training set.
| stat |
A local MCMC algorithm for variable selection with dimension-free mixing
time | Yang et al. (2016) proved that the symmetric random walk Metropolis--Hastings
algorithm for Bayesian variable selection is rapidly mixing under mild
high-dimensional assumptions. In this work, we introduce a novel
Metropolis--Hastings algorithm, which still proposes new states via
add-delete-swap moves but has a much faster mixing time independent of the
number of covariates. The key idea is to use a locally informed proposal scheme
with bounded weights. Motivated by the theoretical analysis of our algorithm,
we further propose a method called "two-stage drift condition" for studying
convergence rates of Markov chains on general state spaces. Simulation studies
show that the new algorithm can efficiently explore the posterior distribution
under various settings. A real data example is also provided, where five
genetic variants associated with cup-to-disk ratio are identified.
| stat |
Escaping Saddle-Points Faster under Interpolation-like Conditions | In this paper, we show that under over-parametrization several standard
stochastic optimization algorithms escape saddle-points and converge to
local-minimizers much faster. One of the fundamental aspects of
over-parametrized models is that they are capable of interpolating the training
data. We show that, under interpolation-like assumptions satisfied by the
stochastic gradients in an over-parametrization setting, the first-order oracle
complexity of Perturbed Stochastic Gradient Descent (PSGD) algorithm to reach
an $\epsilon$-local-minimizer, matches the corresponding deterministic rate of
$\tilde{\mathcal{O}}(1/\epsilon^{2})$. We next analyze Stochastic
Cubic-Regularized Newton (SCRN) algorithm under interpolation-like conditions,
and show that the oracle complexity to reach an $\epsilon$-local-minimizer
under interpolation-like conditions, is
$\tilde{\mathcal{O}}(1/\epsilon^{2.5})$. While this obtained complexity is
better than the corresponding complexity of either PSGD, or SCRN without
interpolation-like assumptions, it does not match the rate of
$\tilde{\mathcal{O}}(1/\epsilon^{1.5})$ corresponding to deterministic
Cubic-Regularized Newton method. It seems further Hessian-based
interpolation-like assumptions are necessary to bridge this gap. We also
discuss the corresponding improved complexities in the zeroth-order settings.
| stat |
Change Point Detection for Nonparametric Regression under Strongly
Mixing Process | In this article, we consider the estimation of the structural change point in
the nonparametric model with dependent observations. We introduce a
maximum-CUSUM-estimation procedure, where the CUSUM statistic is constructed
based on the sum-of-squares aggregation of the difference of the two
Nadaraya-Watson estimates using the observations before and after a specific
time point. Under some mild conditions, we prove that the statistic tends to
zero almost surely if there is no change, and is larger than a threshold
asymptotically almost surely otherwise, which helps us to obtain a
threshold-detection strategy. Furthermore, we demonstrate the strong
consistency of the change point estimator. In the simulation, we discuss the
selection of the bandwidth and the threshold used in the estimation, and show
the robustness of our method in the long-memory scenario. We implement our
method to the data of Nasdaq 100 index and find that the relation between the
realized volatility and the return exhibits several structural changes in
2007-2009.
| stat |
Optimal transport couplings of Gibbs samplers on partitions for unbiased
estimation | Computational couplings of Markov chains provide a practical route to
unbiased Monte Carlo estimation that can utilize parallel computation. However,
these approaches depend crucially on chains meeting after a small number of
transitions. For models that assign data into groups, e.g. mixture models, the
obvious approaches to couple Gibbs samplers fail to meet quickly. This failure
owes to the so-called "label-switching" problem; semantically equivalent
relabelings of the groups contribute well-separated posterior modes that impede
fast mixing and cause large meeting times. We here demonstrate how to avoid
label switching by considering chains as exploring the space of partitions
rather than labelings. Using a metric on this space, we employ an optimal
transport coupling of the Gibbs conditionals. This coupling outperforms
alternative couplings that rely on labelings and, on a real dataset, provides
estimates more precise than usual ergodic averages in the limited time regime.
Code is available at github.com/tinnguyen96/coupling-Gibbs-partition.
| stat |
Knowing what you know: valid and validated confidence sets in multiclass
and multilabel prediction | We develop conformal prediction methods for constructing valid predictive
confidence sets in multiclass and multilabel problems without assumptions on
the data generating distribution. A challenge here is that typical conformal
prediction methods---which give marginal validity (coverage)
guarantees---provide uneven coverage, in that they address easy examples at the
expense of essentially ignoring difficult examples. By leveraging ideas from
quantile regression, we build methods that always guarantee correct coverage
but additionally provide (asymptotically optimal) conditional coverage for both
multiclass and multilabel prediction problems. To address the potential
challenge of exponentially large confidence sets in multilabel prediction, we
build tree-structured classifiers that efficiently account for interactions
between labels. Our methods can be bolted on top of any classification
model---neural network, random forest, boosted tree---to guarantee its
validity. We also provide an empirical evaluation, simultaneously providing new
validation methods, that suggests the more robust coverage of our confidence
sets.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.