title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Canonical Regression Quantiles with application to CEO compensation and
predicting company performance | In using multiple regression methods for prediction, one often considers the
linear combination of explanatory variables as an index. Seeking a single such
index when here are multiple responses is rather more complicated. One
classical approach is to use the coefficients from the leading canonical
correlation. However, methods based on variances are unable to disaggregate
responses by quantile effects, lack robustness, and rely on normal assumptions
for inference. We develop here an alternative regression quantile approach and
apply it to an empirical study of the performance of large publicly held
companies and CEO compensation. The initial results are very promising.
| stat |
A marginal moment matching approach for fitting endemic-epidemic models
to underreported disease surveillance counts | Count data are often subject to underreporting, especially in infectious
disease surveillance. We propose an approximate maximum likelihood method to
fit count time series models from the endemic-epidemic class to underreported
data. The approach is based on marginal moment matching where underreported
processes are approximated through completely observed processes from the same
class. Moreover, the form of the bias when underreporting is ignored or taken
into account via multiplication factors is analysed. Notably, we show that this
leads to a downward bias in model-based estimates of the effective reproductive
number. A marginal moment matching approach can also be used to account for
reporting intervals which are longer than the mean serial interval of a
disease. The good performance of the proposed methodology is demonstrated in
simulation studies. An extension to time-varying parameters and reporting
probabilities is discussed and applied in a case study on weekly rotavirus
gastroenteritis counts in Berlin, Germany.
| stat |
volesti: Volume Approximation and Sampling for Convex Polytopes in R | Sampling from high dimensional distributions and volume approximation of
convex bodies are fundamental operations that appear in optimization, finance,
engineering and machine learning. In this paper we present volesti, a C++
package with an R interface that provides efficient, scalable algorithms for
volume estimation, uniform and Gaussian sampling from convex polytopes. volesti
scales to hundreds of dimensions, handles efficiently three different types of
polyhedra and provides non existing sampling routines to R. We demonstrate the
power of volesti by solving several challenging problems using the R language.
| stat |
Estimating SARS-CoV-2 Infections from Deaths, Confirmed Cases, Tests,
and Random Surveys | There are many sources of data giving information about the number of
SARS-CoV-2 infections in the population, but all have major drawbacks,
including biases and delayed reporting. For example, the number of confirmed
cases largely underestimates the number of infections, deaths lag infections
substantially, while test positivity rates tend to greatly overestimate
prevalence. Representative random prevalence surveys, the only putatively
unbiased source, are sparse in time and space, and the results come with a big
delay. Reliable estimates of population prevalence are necessary for
understanding the spread of the virus and the effects of mitigation strategies.
We develop a simple Bayesian framework to estimate viral prevalence by
combining the main available data sources. It is based on a discrete-time SIR
model with time-varying reproductive parameter. Our model includes likelihood
components that incorporate data of deaths due to the virus, confirmed cases,
and the number of tests administered on each day. We anchor our inference with
data from random sample testing surveys in Indiana and Ohio. We use the results
from these two states to calibrate the model on positive test counts and
proceed to estimate the infection fatality rate and the number of new
infections on each day in each state in the USA. We estimate the extent to
which reported COVID cases have underestimated true infection counts, which was
large, especially in the first months of the pandemic. We explore the
implications of our results for progress towards herd immunity.
| stat |
Linear functional regression with truncated signatures | We place ourselves in a functional regression setting and propose a novel
methodology for regressing a real output on vector-valued functional
covariates. This methodology is based on the notion of signature, which is a
representation of a function as an infinite series of its iterated integrals.
The signature depends crucially on a truncation parameter for which an
estimator is provided, together with theoretical guarantees. An empirical study
on both simulated and real-world datasets shows that the resulting methodology
is competitive with traditional functional linear models, in particular when
the functional covariates take their values in a high dimensional space.
| stat |
colorspace: A Toolbox for Manipulating and Assessing Colors and Palettes | The R package colorspace provides a flexible toolbox for selecting individual
colors or color palettes, manipulating these colors, and employing them in
statistical graphics and data visualizations. In particular, the package
provides a broad range of color palettes based on the HCL
(Hue-Chroma-Luminance) color space. The three HCL dimensions have been shown to
match those of the human visual system very well, thus facilitating intuitive
selection of color palettes through trajectories in this space. Using the HCL
color model general strategies for three types of palettes are implemented: (1)
Qualitative for coding categorical information, i.e., where no particular
ordering of categories is available. (2) Sequential for coding ordered/numeric
information, i.e., going from high to low (or vice versa). (3) Diverging for
coding ordered/numeric information around a central neutral value, i.e., where
colors diverge from neutral to two extremes. To aid selection and application
of these palettes the package also contains scales for use with ggplot2, shiny
(and tcltk) apps for interactive exploration, visualizations of palette
properties, accompanying manipulation utilities (like desaturation and
lighten/darken), and emulation of color vision deficiencies.
| stat |
Handling high correlations in the feature gene selection using
Single-Cell RNA sequencing data | Motivation: Selecting feature genes and predicting cells' phenotype are
typical tasks in the analysis of scRNA-seq data. Many algorithms were developed
for these tasks, but high correlations among genes create challenges
specifically in scRNA-seq analysis, which are not well addressed. Highly
correlated genes lead to unreliable prediction models due to technical
problems, such as multi-collinearity. Most importantly, when a causal gene
(whose variants have a true biological effect on the phenotype) is highly
correlated with other genes, most algorithms select one of them in a
data-driven manner. The correlation structure among genes could change
substantially. Hence, it is critical to build a prediction model based on
causal genes.
Results: To address the issues discussed above, we propose a grouping
algorithm that can be integrated into prediction models. Using real benchmark
scRNA-seq data and simulated cell phenotypes, we show our novel method
significantly outperforms standard models in both prediction and feature
selection. Our algorithm reports the whole group of correlated genes, allowing
researchers to either use their common pattern as a more robust predictor or
conduct follow-up studies to identify the causal genes in the group.
Availability: An R package is being developed and will be available on the
Comprehensive R Archive Network (CRAN) when the paper is published.
| stat |
Matrix-free Penalized Spline Smoothing with Multiple Covariates | The paper motivates high dimensional smoothing with penalized splines and its
numerical calculation in an efficient way. If smoothing is carried out over
three or more covariates the classical tensor product spline bases explode in
their dimension bringing the estimation to its numerical limits. A recent
approach by Siebenborn and Wagner(2019) circumvents storage expensive
implementations by proposing matrix-free calculations which allows to smooth
over several covariates. We extend their approach here by linking penalized
smoothing and its Bayesian formulation as mixed model which provides a
matrix-free calculation of the smoothing parameter to avoid the use of
high-computational cross validation. Further, we show how to extend the ideas
towards generalized regression models. The extended approach is applied to
remote sensing satellite data in combination with spatial smoothing.
| stat |
An Adaptive Approach for Anomaly Detector Selection and Fine-Tuning in
Time Series | The anomaly detection of time series is a hotspot of time series data mining.
The own characteristics of different anomaly detectors determine the abnormal
data that they are good at. There is no detector can be optimizing in all types
of anomalies. Moreover, it still has difficulties in industrial production due
to problems such as a single detector can't be optimized at different time
windows of the same time series. This paper proposes an adaptive model based on
time series characteristics and selecting appropriate detector and run-time
parameters for anomaly detection, which is called ATSDLN(Adaptive Time Series
Detector Learning Network). We take the time series as the input of the model,
and learn the time series representation through FCN. In order to realize the
adaptive selection of detectors and run-time parameters according to the input
time series, the outputs of FCN are the inputs of two sub-networks: the
detector selection network and the run-time parameters selection network. In
addition, the way that the variable layer width design of the parameter
selection sub-network and the introduction of transfer learning make the model
be with more expandability. Through experiments, it is found that ATSDLN can
select appropriate anomaly detector and run-time parameters, and have strong
expandability, which can quickly transfer. We investigate the performance of
ATSDLN in public data sets, our methods outperform other methods in most cases
with higher effect and better adaptation. We also show experimental results on
public data sets to demonstrate how model structure and transfer learning
affect the effectiveness.
| stat |
Simultaneous nonparametric regression in RADWT dictionaries | A new technique for nonparametric regression of multichannel signals is
presented. The technique is based on the use of the Rational-Dilation Wavelet
Transform (RADWT), equipped with a tunable Q-factor able to provide sparse
representations of functions with different oscillations persistence. In
particular, two different frames are obtained by two RADWT with different
Q-factors that give sparse representations of functions with low and high
resonance. It is assumed that the signals are measured simultaneously on
several independent channels and that they share the low resonance component
and the spectral characteristics of the high resonance component. Then, a
regression analysis is performed by means of the grouped lasso penalty.
Furthermore, a result of asymptotic optimality of the estimator is presented
using reasonable assumptions and exploiting recent results on group-lasso like
procedures. Numerical experiments show the performance of the proposed method
in different synthetic scenarios as well as in a real case example for the
analysis and joint detection of sleep spindles and K-complex events for
multiple electroencephalogram (EEG) signals.
| stat |
Empirical Bayes for Large-scale Randomized Experiments: a Spectral
Approach | Large-scale randomized experiments, sometimes called A/B tests, are
increasingly prevalent in many industries. Though such experiments are often
analyzed via frequentist $t$-tests, arguably such analyses are deficient:
$p$-values are hard to interpret and not easily incorporated into
decision-making. As an alternative, we propose an empirical Bayes approach,
which assumes that the treatment effects are realized from a "true prior". This
requires inferring the prior from previous experiments. Following Robbins, we
estimate a family of marginal densities of empirical effects, indexed by the
noise scale. We show that this family is characterized by the heat equation. We
develop a spectral maximum likelihood estimate based on a Fourier series
representation, which can be efficiently computed via convex optimization. In
order to select hyperparameters and compare models, we describe two model
selection criteria. We demonstrate our method on simulated and real data, and
compare posterior inference to that under a Gaussian mixture model of the
prior.
| stat |
A General $\mathcal{O}(n^2)$ Hyper-Parameter Optimization for Gaussian
Process Regression with Cross-Validation and Non-linearly Constrained ADMM | Hyper-parameter optimization remains as the core issue of Gaussian process
(GP) for machine learning nowadays. The benchmark method using maximum
likelihood (ML) estimation and gradient descent (GD) is impractical for
processing big data due to its $O(n^3)$ complexity. Many sophisticated global
or local approximation models, for instance, sparse GP, distributed GP, have
been proposed to address such complexity issue. In this paper, we propose two
novel and general-purpose GP hyper-parameter training schemes (GPCV-ADMM) by
replacing ML with cross-validation (CV) as the fitting criterion and replacing
GD with a non-linearly constrained alternating direction method of multipliers
(ADMM) as the optimization method. The proposed schemes are of $O(n^2)$
complexity for any covariance matrix without special structure. We conduct
various experiments based on both synthetic and real data sets, wherein the
proposed schemes show excellent performance in terms of convergence,
hyper-parameter estimation accuracy, and computational time in comparison with
the traditional ML based routines given in the GPML toolbox.
| stat |
Approximating Lipschitz continuous functions with GroupSort neural
networks | Recent advances in adversarial attacks and Wasserstein GANs have advocated
for use of neural networks with restricted Lipschitz constants. Motivated by
these observations, we study the recently introduced GroupSort neural networks,
with constraints on the weights, and make a theoretical step towards a better
understanding of their expressive power. We show in particular how these
networks can represent any Lipschitz continuous piecewise linear functions. We
also prove that they are well-suited for approximating Lipschitz continuous
functions and exhibit upper bounds on both the depth and size. To conclude, the
efficiency of GroupSort networks compared with more standard ReLU networks is
illustrated in a set of synthetic experiments.
| stat |
Double Robust Semi-Supervised Inference for the Mean: Selection Bias
under MAR Labeling with Decaying Overlap | Semi-supervised (SS) inference has received much attention in recent years.
Apart from a moderate-sized labeled data, L, the SS setting is characterized by
an additional, much larger sized, unlabeled data, U. The setting of |U| >> |L|,
makes SS inference unique and different from the standard missing data
problems, owing to natural violation of the so-called 'positivity' or 'overlap'
assumption. However, most of the SS literature implicitly assumes L and U to be
equally distributed, i.e., no selection bias in the labeling. Inferential
challenges in missing at random (MAR) type labeling allowing for selection
bias, are inevitably exacerbated by the decaying nature of the propensity score
(PS). We address this gap for a prototype problem, the estimation of the
response's mean. We propose a double robust SS (DRSS) mean estimator and give a
complete characterization of its asymptotic properties. The proposed estimator
is consistent as long as either the outcome or the PS model is correctly
specified. When both models are correctly specified, we provide inference
results with a non-standard consistency rate that depends on the smaller size
|L|. The results are also extended to causal inference with imbalanced
treatment groups. Further, we provide several novel choices of models and
estimators of the decaying PS, including a novel offset logistic model and a
stratified labeling model. We present their properties under both high and low
dimensional settings. These may be of independent interest. Lastly, we present
extensive simulations and also a real data application.
| stat |
Learning Energy-Based Models as Generative ConvNets via Multi-grid
Modeling and Sampling | This paper proposes a multi-grid method for learning energy-based generative
ConvNet models of images. For each grid, we learn an energy-based probabilistic
model where the energy function is defined by a bottom-up convolutional neural
network (ConvNet or CNN). Learning such a model requires generating synthesized
examples from the model. Within each iteration of our learning algorithm, for
each observed training image, we generate synthesized images at multiple grids
by initializing the finite-step MCMC sampling from a minimal 1 x 1 version of
the training image. The synthesized image at each subsequent grid is obtained
by a finite-step MCMC initialized from the synthesized image generated at the
previous coarser grid. After obtaining the synthesized examples, the parameters
of the models at multiple grids are updated separately and simultaneously based
on the differences between synthesized and observed examples. We show that this
multi-grid method can learn realistic energy-based generative ConvNet models,
and it outperforms the original contrastive divergence (CD) and persistent CD.
| stat |
Reducing Crowdsourcing to Graphon Estimation, Statistically | Inferring the correct answers to binary tasks based on multiple noisy answers
in an unsupervised manner has emerged as the canonical question for micro-task
crowdsourcing or more generally aggregating opinions. In graphon estimation,
one is interested in estimating edge intensities or probabilities between nodes
using a single snapshot of a graph realization. In the recent literature, there
has been exciting development within both of these topics. In the context of
crowdsourcing, the key intellectual challenge is to understand whether a given
task can be more accurately denoised by aggregating answers collected from
other different tasks. In the context of graphon estimation, precise
information limits and estimation algorithms remain of interest. In this paper,
we utilize a statistical reduction from crowdsourcing to graphon estimation to
advance the state-of-art for both of these challenges. We use concepts from
graphon estimation to design an algorithm that achieves better performance than
the {\em majority voting} scheme for a setup that goes beyond the {\em rank
one} models considered in the literature. We use known explicit lower bounds
for crowdsourcing to provide refined lower bounds for graphon estimation.
| stat |
Localizing differences in smooths with simultaneous confidence bounds on
the true discovery proportion | We demonstrate a method for localizing where two smooths differ using a true
discovery proportion (TDP) based interpretation. The procedure yields a
statement on the proportion of some region where true differences exist between
two smooths, which results from its use of hypothesis tests on groups of basis
coefficients underlying the smooths. The methodology avoids otherwise ad hoc
means of doing so such as performing hypothesis tests on entire smooths of
subsetted data. TDP estimates are 1-$\alpha$ confidence bounded simultaneously,
assuring that the estimate for a region is a lower bound on the proportion of
actual difference, or true discoveries, in that region with high confidence
regardless of the number or location of regions estimated. Our procedure is
based on closed-testing [Hommel, 1986] and recent results of Goeman and Solari
[2011] and Goeman et al [2019]. We develop expressions for the covariance of
quadratic forms because of the multiple regression framework in which we use
the closed-testing procedure, which are shown to be non-negative in many
settings. Our method is well-powered because of a given result on the
off-diagonal decay structure of the covariance matrix of penalized B-splines of
degree two or less. We demonstrate achievement of estimated TDP and nominal
type 1 error rates in simulation and analyze a data set of walking gait of
cerebral palsy patients. Keywords: splines; smoothing; multiple testing;
closed-testing; simultaneous confidence
| stat |
Generalized Matrix Decomposition Regression: Estimation and Inference
for Two-way Structured Data | This paper studies high-dimensional regression with two-way structured data.
To estimate the high-dimensional coefficient vector, we propose the generalized
matrix decomposition regression (GMDR) to efficiently leverage any auxiliary
information on row and column structures. The GMDR extends the principal
component regression (PCR) to two-way structured data, but unlike PCR, the GMDR
selects the components that are most predictive of the outcome, leading to more
accurate prediction. For inference on regression coefficients of individual
variables, we propose the generalized matrix decomposition inference (GMDI), a
general high-dimensional inferential framework for a large family of estimators
that include the proposed GMDR estimator. GMDI provides more flexibility for
modeling relevant auxiliary row and column structures. As a result, GMDI does
not require the true regression coefficients to be sparse; it also allows
dependent and heteroscedastic observations. We study the theoretical properties
of GMDI in terms of both the type-I error rate and power and demonstrate the
effectiveness of GMDR and GMDI on simulation studies and an application to
human microbiome data.
| stat |
Spatial Flow-Field Approximation Using Few Thermodynamic Measurements
Part I: Formulation and Area Averaging | Our investigation raises an important question that is of relevance to the
wider turbomachinery community: how do we estimate the spatial average of a
flow quantity given finite (and sparse) measurements? This paper seeks to
advance efforts to answer this question rigorously. In this paper, we develop a
regularized multivariate linear regression framework for studying engine
temperature measurements. As part of this investigation, we study the
temperature measurements obtained from the same axial plane across five
different engines yielding a total of 82 data-sets. The five different engines
have similar architectures and therefore similar temperature spatial harmonics
are expected. Our problem is to estimate the spatial field in engine
temperature given a few measurements obtained from thermocouples positioned on
a set of rakes. Our motivation for doing so is to understand key engine
temperature modes that cannot be captured in a rig or in computational
simulations, as the cause of these modes may not be replicated in these simpler
environments. To this end, we develop a multivariate linear least squares model
with Tikhonov regularization to estimate the 2D temperature spatial field. Our
model uses a Fourier expansion in the circumferential direction and a quadratic
polynomial expansion in the radial direction. One important component of our
modeling framework is the selection of model parameters, i.e. the harmonics in
the circumferential direction. A training-testing paradigm is proposed and
applied to quantify the harmonics.
| stat |
A Machine Learning Approach for Modelling Parking Duration in Urban
Land-use | Parking is an inevitable issue in the fast-growing developing countries.
Increasing number of vehicles require more and more urban land to be allocated
for parking. However, a little attention has been conferred to the parking
issues in developing countries like India. This study proposes a model for
analysing the influence of car users' socioeconomic and travel characteristics
on parking duration. Specifically, artificial neural networks (ANNs) is
deployed to capture the interrelationship between driver characteristics and
parking duration. ANNs are highly efficient in learning and recognizing
connections between parameters for best prediction of an outcome. Since,
utility of ANNs has been critically limited due to its Black Box nature, the
study involves the use of Garson algorithm and Local interpretable
model-agnostic explanations (LIME) for model interpretations. LIME shows the
prediction for any classification, by approximating it locally with the
developed interpretable model. This study is based on microdata collected
on-site through interview surveys considering two land-uses: office-business
and market/shopping. Results revealed the higher probability of prediction
through LIME and therefore, the methodology can be adopted ubiquitously.
Further, the policy implications are discussed based on the results for both
land-uses. This unique study could lead to enhanced parking policy and
management to achieve the sustainability goals.
| stat |
Model selection criteria of the standard censored regression model based
on the bootstrap sample augmentation mechanism | The statistical regression technique is an extraordinarily essential data
fitting tool to explore the potential possible generation mechanism of the
random phenomenon. Therefore, the model selection or the variable selection is
becoming extremely important so as to identify the most appropriate model with
the most optimal explanation effect on the interesting response. In this paper,
we discuss and compare the bootstrap-based model selection criteria on the
standard censored regression model (Tobit regression model) under the
circumstance of limited observation information. The Monte Carlo numerical
evidence demonstrates that the performances of the model selection criteria
based on the bootstrap sample augmentation strategy will become more
competitive than their alternative ones, such as the Akaike Information
Criterion (AIC) and the Bayesian Information Criterion (BIC) etc. under the
circumstance of the inadequate observation information. Meanwhile, the
numerical simulation experiments further demonstrate that the model
identification risk due to the deficiency of the data information, such as the
high censoring rate and rather limited number of observations, can be
adequately compensated by increasing the scientific computation cost in terms
of the bootstrap sample augmentation strategies. We also apply the recommended
bootstrap-based model selection criterion on the Tobit regression model to fit
the real fidelity dataset.
| stat |
High dimensional multilevel Kriging: A computational mathematics
approach | With the advent of massive data sets much of the computational science and
engineering communities have been moving toward data-driven approaches such as
regression and classification. However, they present a significant challenge
due to the increasing size, complexity and dimensionality of the problems. In
this paper a multilevel Kriging method that scales well with the number of
observations and dimensions is developed. A multilevel basis is constructed
that is adapted to a kD-tree partitioning of the observations. Numerically
unstable covariance matrices with large condition numbers are transformed into
well conditioned multilevel matrices without compromising accuracy. Moreover,
it is shown that the multilevel prediction $exactly$ solves the Best Linear
Unbiased Predictor (BLUP), but is numerically stable. The multilevel method is
tested on numerically unstable problems of up 25 dimensions. Numerical results
show speedups of up to 42,050 for solving the BLUP problem but to the same
accuracy than the traditional iterative approach.
| stat |
Bayesian Optimization for Policy Search via Online-Offline
Experimentation | Online field experiments are the gold-standard way of evaluating changes to
real-world interactive machine learning systems. Yet our ability to explore
complex, multi-dimensional policy spaces - such as those found in
recommendation and ranking problems - is often constrained by the limited
number of experiments that can be run simultaneously. To alleviate these
constraints, we augment online experiments with an offline simulator and apply
multi-task Bayesian optimization to tune live machine learning systems. We
describe practical issues that arise in these types of applications, including
biases that arise from using a simulator and assumptions for the multi-task
kernel. We measure empirical learning curves which show substantial gains from
including data from biased offline experiments, and show how these learning
curves are consistent with theoretical results for multi-task Gaussian process
generalization. We find that improved kernel inference is a significant driver
of multi-task generalization. Finally, we show several examples of Bayesian
optimization efficiently tuning a live machine learning system by combining
offline and online experiments.
| stat |
Efficient Transfer Bayesian Optimization with Auxiliary Information | We propose an efficient transfer Bayesian optimization method, which finds
the maximum of an expensive-to-evaluate black-box function by using data on
related optimization tasks. Our method uses auxiliary information that
represents the task characteristics to effectively transfer knowledge for
estimating a distribution over target functions. In particular, we use a
Gaussian process, in which the mean and covariance functions are modeled with
neural networks that simultaneously take both the auxiliary information and
feature vectors as input. With a neural network mean function, we can estimate
the target function even without evaluations. By using the neural network
covariance function, we can extract nonlinear correlation among feature vectors
that are shared across related tasks. Our Gaussian process-based formulation
not only enables an analytic calculation of the posterior distribution but also
swiftly adapts the target function to observations. Our method is also
advantageous because the computational costs scale linearly with the number of
source tasks. Through experiments using a synthetic dataset and datasets for
finding the optimal pedestrian traffic regulations and optimal machine learning
algorithms, we demonstrate that our method identifies the optimal points with
fewer target function evaluations than existing methods.
| stat |
Understanding and Quantifying Adversarial Examples Existence in Linear
Classification | State-of-art deep neural networks (DNN) are vulnerable to attacks by
adversarial examples: a carefully designed small perturbation to the input,
that is imperceptible to human, can mislead DNN. To understand the root cause
of adversarial examples, we quantify the probability of adversarial example
existence for linear classifiers. Previous mathematical definition of
adversarial examples only involves the overall perturbation amount, and we
propose a more practical relevant definition of strong adversarial examples
that separately limits the perturbation along the signal direction also. We
show that linear classifiers can be made robust to strong adversarial examples
attack in cases where no adversarial robust linear classifiers exist under the
previous definition. The quantitative formulas are confirmed by numerical
experiments using a linear support vector machine (SVM) classifier. The results
suggest that designing general strong-adversarial-robust learning systems is
feasible but only through incorporating human knowledge of the underlying
classification problem.
| stat |
Making Recursive Bayesian Inference Accessible | Bayesian models provide recursive inference naturally because they can
formally reconcile new data and existing scientific information. However,
popular use of Bayesian methods often avoids priors that are based on exact
posterior distributions resulting from former studies. Two existing Recursive
Bayesian methods are: Prior- and Proposal-Recursive Bayes. Prior-Recursive
Bayes uses Bayesian updating, fitting models to partitions of data
sequentially, and provides a way to accommodate new data as they become
available using the posterior from the previous stage as the prior in the new
stage based on the latest data. Proposal-Recursive Bayes is intended for use
with hierarchical Bayesian models and uses a set of transient priors in first
stage independent analyses of the data partitions. The second stage of
Proposal-Recursive Bayes uses the posteriors from the first stage as proposals
in an MCMC algorithm to fit the full model. We combine Prior- and
Proposal-Recursive concepts to fit any Bayesian model, and often with
computational improvements. We demonstrate our method with two case studies.
Our approach has implications for big data, streaming data, and optimal
adaptive design situations.
| stat |
Bias-Variance Tradeoff in a Sliding Window Implementation of the
Stochastic Gradient Algorithm | This paper provides a framework to analyze stochastic gradient algorithms in
a mean squared error (MSE) sense using the asymptotic normality result of the
stochastic gradient descent (SGD) iterates. We perform this analysis by taking
the asymptotic normality result and applying it to the finite iteration case.
Specifically, we look at problems where the gradient estimators are biased and
have reduced variance and compare the iterates generated by these gradient
estimators to the iterates generated by the SGD algorithm. We use the work of
Fabian to characterize the mean and the variance of the distribution of the
iterates in terms of the bias and the covariance matrix of the gradient
estimators. We introduce the sliding window SGD (SW-SGD) algorithm, with its
proof of convergence, which incurs a lower MSE than the SGD algorithm on
quadratic and convex problems. Lastly, we present some numerical results to
show the effectiveness of this framework and the superiority of SW-SGD
algorithm over the SGD algorithm.
| stat |
Super-efficiency of automatic differentiation for functions defined as a
minimum | In min-min optimization or max-min optimization, one has to compute the
gradient of a function defined as a minimum. In most cases, the minimum has no
closed-form, and an approximation is obtained via an iterative algorithm. There
are two usual ways of estimating the gradient of the function: using either an
analytic formula obtained by assuming exactness of the approximation, or
automatic differentiation through the algorithm. In this paper, we study the
asymptotic error made by these estimators as a function of the optimization
error. We find that the error of the automatic estimator is close to the square
of the error of the analytic estimator, reflecting a super-efficiency
phenomenon. The convergence of the automatic estimator greatly depends on the
convergence of the Jacobian of the algorithm. We analyze it for gradient
descent and stochastic gradient descent and derive convergence rates for the
estimators in these cases. Our analysis is backed by numerical experiments on
toy problems and on Wasserstein barycenter computation. Finally, we discuss the
computational complexity of these estimators and give practical guidelines to
chose between them.
| stat |
Coupling techniques for nonlinear ensemble filtering | We consider filtering in high-dimensional non-Gaussian state-space models
with intractable transition kernels, nonlinear and possibly chaotic dynamics,
and sparse observations in space and time. We propose a novel filtering
methodology that harnesses transportation of measures, convex optimization, and
ideas from probabilistic graphical models to yield robust ensemble
approximations of the filtering distribution in high dimensions. Our approach
can be understood as the natural generalization of the ensemble Kalman filter
(EnKF) to nonlinear updates, using stochastic or deterministic couplings. The
use of nonlinear updates can reduce the intrinsic bias of the EnKF at a
marginal increase in computational cost. We avoid any form of importance
sampling and introduce non-Gaussian localization approaches for dimension
scalability. Our framework achieves state-of-the-art tracking performance on
challenging configurations of the Lorenz-96 model in the chaotic regime.
| stat |
A general framework for Vecchia approximations of Gaussian processes | Gaussian processes (GPs) are commonly used as models for functions, time
series, and spatial fields, but they are computationally infeasible for large
datasets. Focusing on the typical setting of modeling data as a GP plus an
additive noise term, we propose a generalization of the Vecchia (1988) approach
as a framework for GP approximations. We show that our general Vecchia approach
contains many popular existing GP approximations as special cases, allowing for
comparisons among the different methods within a unified framework.
Representing the models by directed acyclic graphs, we determine the sparsity
of the matrices necessary for inference, which leads to new insights regarding
the computational properties. Based on these results, we propose a novel sparse
general Vecchia approximation, which ensures computational feasibility for
large spatial datasets but can lead to considerable improvements in
approximation accuracy over Vecchia's original approach. We provide several
theoretical results and conduct numerical comparisons. We conclude with
guidelines for the use of Vecchia approximations in spatial statistics.
| stat |
Data Shapley: Equitable Valuation of Data for Machine Learning | As data becomes the fuel driving technological and economic growth, a
fundamental challenge is how to quantify the value of data in algorithmic
predictions and decisions. For example, in healthcare and consumer markets, it
has been suggested that individuals should be compensated for the data that
they generate, but it is not clear what is an equitable valuation for
individual data. In this work, we develop a principled framework to address
data valuation in the context of supervised machine learning. Given a learning
algorithm trained on $n$ data points to produce a predictor, we propose data
Shapley as a metric to quantify the value of each training datum to the
predictor performance. Data Shapley value uniquely satisfies several natural
properties of equitable data valuation. We develop Monte Carlo and
gradient-based methods to efficiently estimate data Shapley values in practical
settings where complex learning algorithms, including neural networks, are
trained on large datasets. In addition to being equitable, extensive
experiments across biomedical, image and synthetic data demonstrate that data
Shapley has several other benefits: 1) it is more powerful than the popular
leave-one-out or leverage score in providing insight on what data is more
valuable for a given learning task; 2) low Shapley value data effectively
capture outliers and corruptions; 3) high Shapley value data inform what type
of new data to acquire to improve the predictor.
| stat |
A Latent Topic Model with Markovian Transition for Process Data | We propose a latent topic model with a Markovian transition for process data,
which consist of time-stamped events recorded in a log file. Such data are
becoming more widely available in computer-based educational assessment with
complex problem solving items. The proposed model can be viewed as an extension
of the hierarchical Bayesian topic model with a hidden Markov structure to
accommodate the underlying evolution of an examinee's latent state. Using topic
transition probabilities along with response times enables us to capture
examinees' learning trajectories, making clustering/classification more
efficient. A forward-backward variational expectation-maximization (FB-VEM)
algorithm is developed to tackle the challenging computational problem. Useful
theoretical properties are established under certain asymptotic regimes. The
proposed method is applied to a complex problem solving item in 2012 Programme
for International Student Assessment (PISA 2012).
| stat |
COPOD: Copula-Based Outlier Detection | Outlier detection refers to the identification of rare items that are deviant
from the general data distribution. Existing approaches suffer from high
computational complexity, low predictive capability, and limited
interpretability. As a remedy, we present a novel outlier detection algorithm
called COPOD, which is inspired by copulas for modeling multivariate data
distribution. COPOD first constructs an empirical copula, and then uses it to
predict tail probabilities of each given data point to determine its level of
"extremeness". Intuitively, we think of this as calculating an anomalous
p-value. This makes COPOD both parameter-free, highly interpretable, and
computationally efficient. In this work, we make three key contributions, 1)
propose a novel, parameter-free outlier detection algorithm with both great
performance and interpretability, 2) perform extensive experiments on 30
benchmark datasets to show that COPOD outperforms in most cases and is also one
of the fastest algorithms, and 3) release an easy-to-use Python implementation
for reproducibility.
| stat |
On Tree-based Methods for Similarity Learning | In many situations, the choice of an adequate similarity measure or metric on
the feature space dramatically determines the performance of machine learning
methods. Building automatically such measures is the specific purpose of
metric/similarity learning. In Vogel et al. (2018), similarity learning is
formulated as a pairwise bipartite ranking problem: ideally, the larger the
probability that two observations in the feature space belong to the same class
(or share the same label), the higher the similarity measure between them. From
this perspective, the ROC curve is an appropriate performance criterion and it
is the goal of this article to extend recursive tree-based ROC optimization
techniques in order to propose efficient similarity learning algorithms. The
validity of such iterative partitioning procedures in the pairwise setting is
established by means of results pertaining to the theory of U-processes and
from a practical angle, it is discussed at length how to implement them by
means of splitting rules specifically tailored to the similarity learning task.
Beyond these theoretical/methodological contributions, numerical experiments
are displayed and provide strong empirical evidence of the performance of the
algorithmic approaches we propose.
| stat |
Data science is science's second chance to get causal inference right: A
classification of data science tasks | Causal inference from observational data is the goal of many data analyses in
the health and social sciences. However, academic statistics has often frowned
upon data analyses with a causal objective. The introduction of the term "data
science" provides a historic opportunity to redefine data analysis in such a
way that it naturally accommodates causal inference from observational data.
Like others before, we organize the scientific contributions of data science
into three classes of tasks: Description, prediction, and counterfactual
prediction (which includes causal inference). An explicit classification of
data science tasks is necessary to discuss the data, assumptions, and analytics
required to successfully accomplish each task. We argue that a failure to
adequately describe the role of subject-matter expert knowledge in data
analysis is a source of widespread misunderstandings about data science.
Specifically, causal analyses typically require not only good data and
algorithms, but also domain expert knowledge. We discuss the implications for
the use of data science to guide decision-making in the real world and to train
data scientists.
| stat |
Modeling of Daily Precipitation Amounts Using the Mixed Gamma Weibull
Distribution | By recognizing that the main difficulty of the modeling of daily
precipitation amounts is the selection of an appropriate probability
distribution, this study aims to establish a model selection framework to
identify the appropriate probability distribution for the modeling of daily
precipitation amounts from the commonly used probability distributions, i.e.
the exponential, Gamma, Weibull, and mixed exponential distributions. The mixed
Gamma Weibull (MGW) distribution serves this purpose because all the commonly
used probability distributions are special cases of the MGW distribution, and
the MGW distribution integrates all the commonly used probability distributions
into one framework. Finally, via the large sample inference of likelihood
ratios, a model selection criterion can be established to identify the
appropriate model for the modeling of daily precipitation amounts from the MGW
distribution framework.
| stat |
Neural Architecture Search: A Survey | Deep Learning has enabled remarkable progress over the last years on a
variety of tasks, such as image recognition, speech recognition, and machine
translation. One crucial aspect for this progress are novel neural
architectures. Currently employed architectures have mostly been developed
manually by human experts, which is a time-consuming and error-prone process.
Because of this, there is growing interest in automated neural architecture
search methods. We provide an overview of existing work in this field of
research and categorize them according to three dimensions: search space,
search strategy, and performance estimation strategy.
| stat |
A review of Gaussian Markov models for conditional independence | Markov models lie at the interface between statistical independence in a
probability distribution and graph separation properties. We review model
selection and estimation in directed and undirected Markov models with Gaussian
parametrization, emphasizing the main similarities and differences. These two
model classes are similar but not equivalent, although they share a common
intersection. We present the existing results from a historical perspective,
taking into account the amount of literature existing from both the artificial
intelligence and statistics research communities, where these models were
originated. We cover classical topics such as maximum likelihood estimation and
model selection via hypothesis testing, but also more modern approaches like
regularization and Bayesian methods. We also discuss how the Markov models
reviewed fit in the rich hierarchy of other, higher level Markov model classes.
Finally, we close the paper overviewing relaxations of the Gaussian assumption
and pointing out the main areas of application where these Markov models are
nowadays used.
| stat |
Sequential Bayesian Experimental Design for Implicit Models via Mutual
Information | Bayesian experimental design (BED) is a framework that uses statistical
models and decision making under uncertainty to optimise the cost and
performance of a scientific experiment. Sequential BED, as opposed to static
BED, considers the scenario where we can sequentially update our beliefs about
the model parameters through data gathered in the experiment. A class of models
of particular interest for the natural and medical sciences are implicit
models, where the data generating distribution is intractable, but sampling
from it is possible. Even though there has been a lot of work on static BED for
implicit models in the past few years, the notoriously difficult problem of
sequential BED for implicit models has barely been touched upon. We address
this gap in the literature by devising a novel sequential design framework for
parameter estimation that uses the Mutual Information (MI) between model
parameters and simulated data as a utility function to find optimal
experimental designs, which has not been done before for implicit models. Our
approach uses likelihood-free inference by ratio estimation to simultaneously
estimate posterior distributions and the MI. During the sequential BED
procedure we utilise Bayesian optimisation to help us optimise the MI utility.
We find that our framework is efficient for the various implicit models tested,
yielding accurate parameter estimates after only a few iterations.
| stat |
Variable Selection with ABC Bayesian Forests | Few problems in statistics are as perplexing as variable selection in the
presence of very many redundant covariates. The variable selection problem is
most familiar in parametric environments such as the linear model or additive
variants thereof. In this work, we abandon the linear model framework, which
can be quite detrimental when the covariates impact the outcome in a non-linear
way, and turn to tree-based methods for variable selection. Such variable
screening is traditionally done by pruning down large trees or by ranking
variables based on some importance measure. Despite heavily used in practice,
these ad-hoc selection rules are not yet well understood from a theoretical
point of view. In this work, we devise a Bayesian tree-based probabilistic
method and show that it is consistent for variable selection when the
regression surface is a smooth mix of $p>n$ covariates. These results are the
first model selection consistency results for Bayesian forest priors.
Probabilistic assessment of variable importance is made feasible by a
spike-and-slab wrapper around sum-of-trees priors. Sampling from posterior
distributions over trees is inherently very difficult. As an alternative to
MCMC, we propose ABC Bayesian Forests, a new ABC sampling method based on
data-splitting that achieves higher ABC acceptance rate. We show that the
method is robust and successful at finding variables with high marginal
inclusion probabilities. Our ABC algorithm provides a new avenue towards
approximating the median probability model in non-parametric setups where the
marginal likelihood is intractable.
| stat |
A generalised and fully Bayesian framework for ensemble updating | We propose a generalised framework for the updating of a prior ensemble to a
posterior ensemble, an essential yet challenging part in ensemble-based
filtering methods. The proposed framework is based on a generalised and fully
Bayesian view on the traditional ensemble Kalman filter (EnKF). In the EnKF,
the updating of the ensemble is based on Gaussian assumptions, whereas in our
general setup the updating may be based on another parametric family. In
addition, we propose to formulate an optimality criterion and to find the
optimal update with respect to this criterion. The framework is fully Bayesian
in the sense that the parameters of the assumed forecast model are treated as
random variables. As a consequence, a parameter vector is simulated, for each
ensemble member, prior to the updating. In contrast to existing fully Bayesian
approaches, where the parameters are simulated conditionally on all the
forecast samples, the parameters are in our framework simulated conditionally
on both the data and all the forecast samples, except the forecast sample which
is to be updated. The proposed framework is studied in detail for two
parametric families: the linear-Gaussian model and the finite state-space
hidden Markov model. For both cases, we present simulation examples and compare
the results with existing ensemble-based filtering methods. The results of the
proposed approach indicate a promising performance. In particular, the filter
based on the linear-Gaussian model gives a more realistic representation of the
uncertainty than the traditional EnKF, and the effect of not conditioning on
the forecast sample which is to be updated when simulating the parameters is
remarkable.
| stat |
Integrative Density Forecast and Uncertainty Quantification of Wind
Power Generation | The volatile nature of wind power generation creates challenges in achieving
secure power grid operations. It is, therefore, necessary to make accurate wind
power prediction and its uncertainty quantification. Wind power forecasting
usually depends on wind speed prediction and the wind-to-power conversion
process. However, most current wind power prediction models only consider
portions of the uncertainty. This paper develops an integrative framework for
predicting wind power density, considering uncertainties arising from both wind
speed prediction and the wind-to-power conversion process. Specifically, we
model wind speed using the inhomogeneous Geometric Brownian Motion and convert
the wind speed prediction density into the wind power density in a closed-form.
The resulting wind power density allows quantifying prediction uncertainties
through prediction intervals. To forecast the power output, we minimize the
expected prediction cost with (unequal) penalties on the overestimation and
underestimation. We show the predictive power of the proposed approach using
data from multiple operating wind farms located at different sites.
| stat |
A Distributed and Integrated Method of Moments for High-Dimensional
Correlated Data Analysis | This paper is motivated by a regression analysis of electroencephalography
(EEG) neuroimaging data with high-dimensional correlated responses with
multi-level nested correlations. We develop a divide-and-conquer procedure
implemented in a fully distributed and parallelized computational scheme for
statistical estimation and inference of regression parameters. Despite
significant efforts in the literature, the computational bottleneck associated
with high-dimensional likelihoods prevents the scalability of existing methods.
The proposed method addresses this challenge by dividing responses into
subvectors to be analyzed separately and in parallel on a distributed platform
using pairwise composite likelihood. Theoretical challenges related to
combining results from dependent data are overcome in a statistically efficient
way using a meta-estimator derived from Hansen's generalized method of moments.
We provide a rigorous theoretical framework for efficient estimation,
inference, and goodness-of-fit tests. We develop an R package for ease of
implementation. We illustrate our method's performance with simulations and the
analysis of the EEG data, and find that iron deficiency is significantly
associated with two auditory recognition memory related potentials in the left
parietal-occipital region of the brain.
| stat |
Special Function Methods for Bursty Models of Transcription | We explore a Markov model used in the analysis of gene expression, involving
the bursty production of pre-mRNA, its conversion to mature mRNA, and its
consequent degradation. We demonstrate that the integration used to compute the
solution of the stochastic system can be approximated by the evaluation of
special functions. Furthermore, the form of the special function solution
generalizes to a broader class of burst distributions. In light of the broader
goal of biophysical parameter inference from transcriptomics data, we apply the
method to simulated data, demonstrating effective control of precision and
runtime. Finally, we suggest a non-Bayesian approach to reducing the
computational complexity of parameter inference to linear order in state space
size and number of candidate parameters.
| stat |
Asymptotic Analysis via Stochastic Differential Equations of Gradient
Descent Algorithms in Statistical and Computational Paradigms | This paper investigates asymptotic behaviors of gradient descent algorithms
(particularly accelerated gradient descent and stochastic gradient descent) in
the context of stochastic optimization arising in statistics and machine
learning where objective functions are estimated from available data. We show
that these algorithms can be computationally modeled by continuous-time
ordinary or stochastic differential equations. We establish gradient flow
central limit theorems to describe the limiting dynamic behaviors of these
computational algorithms and the large-sample performances of the related
statistical procedures, as the number of algorithm iterations and data size
both go to infinity, where the gradient flow central limit theorems are
governed by some linear ordinary or stochastic differential equations like
time-dependent Ornstein-Uhlenbeck processes. We illustrate that our study can
provide a novel unified framework for a joint computational and statistical
asymptotic analysis, where the computational asymptotic analysis studies
dynamic behaviors of these algorithms with the time (or the number of
iterations in the algorithms), the statistical asymptotic analysis investigates
large sample behaviors of the statistical procedures (like estimators and
classifiers) that the algorithms are applied to compute, and in fact the
statistical procedures are equal to the limits of the random sequences
generated from these iterative algorithms as the number of iterations goes to
infinity. The joint analysis results based on the obtained gradient flow
central limit theorems can identify four factors - learning rate, batch size,
gradient covariance, and Hessian - to derive new theory regarding the local
minima found by stochastic gradient descent for solving non-convex optimization
problems.
| stat |
Adjusted composite likelihood for robust Bayesian meta-analysis | A composite likelihood is a non-genuine likelihood function that allows to
make inference on limited aspects of a model, such as marginal or conditional
distributions. Composite likelihoods are not proper likelihoods and need
therefore calibration for their use in inference, from both a frequentist and a
Bayesian perspective. The maximizer to the composite likelihood can serve as an
estimator and its variance is assessed by means of a suitably defined sandwich
matrix. In the Bayesian setting, the composite likelihood can be adjusted by
means of magnitude and curvature methods. Magnitude methods imply raising the
likelihood to a constant, while curvature methods imply evaluating the
likelihood at a different point by translating, rescaling and rotating the
parameter vector. Some authors argue that curvature methods are more reliable
in general, but others proved that magnitude methods are sufficient to recover,
for instance, the null distribution of a test statistic. We propose a simple
calibration for the marginal posterior distribution of a scalar parameter of
interest which is invariant to monotonic and smooth transformations. This can
be enough for instance in medical statistics, where a single scalar effect
measure is often the target.
| stat |
An ensemble approach to short-term forecast of COVID-19 intensive care
occupancy in Italian Regions | The availability of intensive care beds during the Covid-19 epidemic is
crucial to guarantee the best possible treatment to severely affected patients.
In this work we show a simple strategy for short-term prediction of Covid-19
ICU beds, that has proved very effective during the Italian outbreak in
February to May 2020. Our approach is based on an optimal ensemble of two
simple methods: a generalized linear mixed regression model which pools
information over different areas, and an area-specific non-stationary integer
autoregressive methodology. Optimal weights are estimated using a
leave-last-out rationale. The approach has been set up and validated during the
epidemic in Italy. A report of its performance for predicting ICU occupancy at
Regional level is included.
| stat |
Bayesian inference for network Poisson models | This work is motivated by the analysis of ecological interaction networks.
Poisson stochastic blockmodels are widely used in this field to decipher the
structure that underlies a weighted network, while accounting for covariate
effects. Efficient algorithms based on variational approximations exist for
frequentist inference, but without statistical guaranties as for the resulting
estimates. In absence of variational Bayes estimates, we show that a good proxy
of the posterior distribution can be straightforwardly derived from the
frequentist variational estimation procedure, using a Laplace approximation. We
use this proxy to sample from the true posterior distribution via a sequential
Monte-Carlo algorithm. As shown in the simulation study, the efficiency of the
posterior sampling is greatly improved by the accuracy of the approximate
posterior distribution. The proposed procedure can be easily extended to other
latent variable models. We use this methodology to assess the influence of
available covariates on the organization of two ecological networks, as well as
the existence of a residual interaction structure.
| stat |
Hyperparameter Learning via Distributional Transfer | Bayesian optimisation is a popular technique for hyperparameter learning but
typically requires initial exploration even in cases where similar prior tasks
have been solved. We propose to transfer information across tasks using learnt
representations of training datasets used in those tasks. This results in a
joint Gaussian process model on hyperparameters and data representations.
Representations make use of the framework of distribution embeddings into
reproducing kernel Hilbert spaces. The developed method has a faster
convergence compared to existing baselines, in some cases requiring only a few
evaluations of the target objective.
| stat |
Imperceptible Adversarial Attacks on Tabular Data | Security of machine learning models is a concern as they may face adversarial
attacks for unwarranted advantageous decisions. While research on the topic has
mainly been focusing on the image domain, numerous industrial applications, in
particular in finance, rely on standard tabular data. In this paper, we discuss
the notion of adversarial examples in the tabular domain. We propose a
formalization based on the imperceptibility of attacks in the tabular domain
leading to an approach to generate imperceptible adversarial examples.
Experiments show that we can generate imperceptible adversarial examples with a
high fooling rate.
| stat |
On statistical deficiency: Why the test statistic of the matching method
is hopelessly underpowered and uniquely informative | The random variate m is, in combinatorics, a basis for comparing
permutations, as well as the solution to a centuries-old riddle involving the
mishandling of hats. In statistics, m is the test statistic for a disused null
hypothesis statistical test (NHST) of association, the matching method. In this
paper, I show that the matching method has an absolute and relatively low limit
on its statistical power. I do so first by reinterpreting Rae's theorem, which
describes the joint distributions of m with several rank correlation statistics
under a true null. I then derive this property solely from m's unconditional
sampling distribution, on which basis I develop the concept of a deficient
statistic: a statistic that is insufficient and inconsistent and inefficient
with respect to its parameter. Finally, I demonstrate an application for m that
makes use of its deficiency to qualify the sampling error in a jointly
estimated sample correlation.
| stat |
Fast Approximate Geodesics for Deep Generative Models | The length of the geodesic between two data points along a Riemannian
manifold, induced by a deep generative model, yields a principled measure of
similarity. Current approaches are limited to low-dimensional latent spaces,
due to the computational complexity of solving a non-convex optimisation
problem. We propose finding shortest paths in a finite graph of samples from
the aggregate approximate posterior, that can be solved exactly, at greatly
reduced runtime, and without a notable loss in quality. Our approach,
therefore, is hence applicable to high-dimensional problems, e.g., in the
visual domain. We validate our approach empirically on a series of experiments
using variational autoencoders applied to image data, including the Chair,
FashionMNIST, and human movement data sets.
| stat |
Bayesian Model Selection on Random Networks | A general Bayesian framework for model selection on random network models
regarding their features is considered. The goal is to develop a principle
Bayesian model selection approach to compare different fittable, not
necessarily nested, models for inference on those network realisations. The
criterion for random network models regarding the comparison is formulated via
Bayes factors and penalizing using the mostwidely used loss functions.
Parametrizations are different in different spaces. To overcome this problem we
incorporate and encode different aspects of complexities in terms of observable
spaces. Thus, given a range of values for a feature, network realisationsare
extracted. The proposed principle approach is based on finding random network
models, such that a reasonable trade off between the interested feature and the
complexity of the model is preserved, avoiding overfitting problems.
| stat |
Likelihood-based Inference for Partially Observed Epidemics on Dynamic
Networks | We propose a generative model and an inference scheme for epidemic processes
on dynamic, adaptive contact networks. Network evolution is formulated as a
link-Markovian process, which is then coupled to an individual-level stochastic
SIR model, in order to describe the interplay between epidemic dynamics on a
network and network link changes. A Markov chain Monte Carlo framework is
developed for likelihood-based inference from partial epidemic observations,
with a novel data augmentation algorithm specifically designed to deal with
missing individual recovery times under the dynamic network setting. Through a
series of simulation experiments, we demonstrate the validity and flexibility
of the model as well as the efficacy and efficiency of the data augmentation
inference scheme. The model is also applied to a recent real-world dataset on
influenza-like-illness transmission with high-resolution social contact
tracking records.
| stat |
Optimally estimating the sample standard deviation from the five-number
summary | When reporting the results of clinical studies, some researchers may choose
the five-number summary (including the sample median, the first and third
quartiles, and the minimum and maximum values) rather than the sample mean and
standard deviation, particularly for skewed data. For these studies, when
included in a meta-analysis, it is often desired to convert the five-number
summary back to the sample mean and standard deviation. For this purpose,
several methods have been proposed in the recent literature and they are
increasingly used nowadays. In this paper, we propose to further advance the
literature by developing a smoothly weighted estimator for the sample standard
deviation that fully utilizes the sample size information. For ease of
implementation, we also derive an approximation formula for the optimal weight,
as well as a shortcut formula for the sample standard deviation. Numerical
results show that our new estimator provides a more accurate estimate for
normal data and also performs favorably for non-normal data. Together with the
optimal sample mean estimator in Luo et al., our new methods have dramatically
improved the existing methods for data transformation, and they are capable to
serve as "rules of thumb" in meta-analysis for studies reported with the
five-number summary. Finally for practical use, an Excel spreadsheet and an
online calculator are also provided for implementing our optimal estimators.
| stat |
Towards Optimal Algorithms for Multi-Player Bandits without Collision
Sensing Information | We propose a novel algorithm for multi-player multi-armed bandits without
collision sensing information. Our algorithm circumvents two problems shared by
all state-of-the-art algorithms: it does not need as an input a lower bound on
the minimal expected reward of an arm, and its performance does not scale
inversely proportionally to the minimal expected reward. We prove a theoretical
regret upper bound to justify these claims. We complement our theoretical
results with numerical experiments, showing that the proposed algorithm
outperforms state-of-the-art in practice as well.
| stat |
Estimating Density Models with Truncation Boundaries using Score
Matching | Truncated densities are probability density functions defined on truncated
domains. They share the same parametric form with their non-truncated
counterparts up to a normalizing constant. Since the computation of their
normalizing constants is usually infeasible, Maximum Likelihood Estimation
cannot be easily applied to estimate truncated density models. Score Matching
(SM) is a powerful tool for fitting parameters using only unnormalized models.
However, it cannot be directly applied here as boundary conditions used to
derive a tractable SM objective are not satisfied by truncated densities. In
this paper, we study parameter estimation for truncated probability densities
using SM. The estimator minimizes a weighted Fisher divergence. The weight
function is simply the shortest distance from the data point to the boundary of
the domain. We show this choice of weight function naturally arises from
minimizing the Stein discrepancy as well as upperbounding the finite-sample
statistical estimation error. The usefulness of our method is demonstrated by
numerical experiments and a Chicago crime dataset. We also show that proposed
density estimation can correct the outlier-trimming bias caused by aggressive
outlier detection methods.
| stat |
Block-wise Minimization-Majorization algorithm for Huber's criterion:
sparse learning and applications | Huber's criterion can be used for robust joint estimation of regression and
scale parameters in the linear model. Huber's (Huber, 1981) motivation for
introducing the criterion stemmed from non-convexity of the joint maximum
likelihood objective function as well as non-robustness (unbounded influence
function) of the associated ML-estimate of scale. In this paper, we illustrate
how the original algorithm proposed by Huber can be set within the block-wise
minimization majorization framework. In addition, we propose novel
data-adaptive step sizes for both the location and scale, which are further
improving the convergence. We then illustrate how Huber's criterion can be used
for sparse learning of underdetermined linear model using the iterative hard
thresholding approach. We illustrate the usefulness of the algorithms in an
image denoising application and simulation studies.
| stat |
Flexible Bayesian Nonlinear Model Configuration | Regression models are used in a wide range of applications providing a
powerful scientific tool for researchers from different fields. Linear models
are often not sufficient to describe the complex relationship between input
variables and a response. This relationship can be better described by
non-linearities and complex functional interactions. Deep learning models have
been extremely successful in terms of prediction although they are often
difficult to specify and potentially suffer from overfitting. In this paper, we
introduce a class of Bayesian generalized nonlinear regression models with a
comprehensive non-linear feature space. Non-linear features are generated
hierarchically, similarly to deep learning, but have additional flexibility on
the possible types of features to be considered. This flexibility, combined
with variable selection, allows us to find a small set of important features
and thereby more interpretable models. A genetically modified Markov chain
Monte Carlo algorithm is developed to make inference. Model averaging is also
possible within our framework. In various applications, we illustrate how our
approach is used to obtain meaningful non-linear models. Additionally, we
compare its predictive performance with a number of machine learning
algorithms.
| stat |
A Goodness-of-Fit Test for Statistical Models | Statistical modeling plays a fundamental role in understanding the underlying
mechanism of massive data (statistical inference) and predicting the future
(statistical prediction). Although all models are wrong, researchers try their
best to make some of them be useful. The question here is how can we measure
the usefulness of a statistical model for the data in hand? This is key to
statistical prediction. The important statistical problem of testing whether
the observations follow the proposed statistical model has only attracted
relatively few attentions. In this paper, we proposed a new framework for this
problem through building its connection with two-sample distribution
comparison. The proposed method can be applied to evaluate a wide range of
models. Examples are given to show the performance of the proposed method.
| stat |
Fast mixing of Metropolized Hamiltonian Monte Carlo: Benefits of
multi-step gradients | Hamiltonian Monte Carlo (HMC) is a state-of-the-art Markov chain Monte Carlo
sampling algorithm for drawing samples from smooth probability densities over
continuous spaces. We study the variant most widely used in practice,
Metropolized HMC with the St\"{o}rmer-Verlet or leapfrog integrator, and make
two primary contributions. First, we provide a non-asymptotic upper bound on
the mixing time of the Metropolized HMC with explicit choices of step-size and
number of leapfrog steps. This bound gives a precise quantification of the
faster convergence of Metropolized HMC relative to simpler MCMC algorithms such
as the Metropolized random walk, or Metropolized Langevin algorithm. Second, we
provide a general framework for sharpening mixing time bounds of Markov chains
initialized at a substantial distance from the target distribution over
continuous spaces. We apply this sharpening device to the Metropolized random
walk and Langevin algorithms, thereby obtaining improved mixing time bounds
from a non-warm initial distribution.
| stat |
Dimensionality reduction, regularization, and generalization in
overparameterized regressions | Overparameterization in deep learning is powerful: Very large models fit the
training data perfectly and yet generalize well. This realization brought back
the study of linear models for regression, including ordinary least squares
(OLS), which, like deep learning, shows a "double descent" behavior. This
involves two features: (1) The risk (out-of-sample prediction error) can grow
arbitrarily when the number of samples $n$ approaches the number of parameters
$p$, and (2) the risk decreases with $p$ at $p>n$, sometimes achieving a lower
value than the lowest risk at $p<n$. The divergence of the risk for OLS at
$p\approx n$ is related to the condition number of the empirical covariance in
the feature set. For this reason, it can be avoided with regularization. In
this work we show that it can also be avoided with a PCA-based dimensionality
reduction. We provide a finite upper bound for the risk of the PCA-based
estimator. This result is in contrast to recent work that shows that a
different form of dimensionality reduction -- one based on the population
covariance instead of the empirical covariance -- does not avoid the
divergence. We connect these results to an analysis of adversarial attacks,
which become more effective as they raise the condition number of the empirical
covariance of the features. We show that OLS is arbitrarily susceptible to
data-poisoning attacks in the overparameterized regime -- unlike the
underparameterized regime -- and that regularization and dimensionality
reduction improve the robustness.
| stat |
Towards a mathematical framework to inform Neural Network modelling via
Polynomial Regression | Even when neural networks are widely used in a large number of applications,
they are still considered as black boxes and present some difficulties for
dimensioning or evaluating their prediction error. This has led to an
increasing interest in the overlapping area between neural networks and more
traditional statistical methods, which can help overcome those problems. In
this article, a mathematical framework relating neural networks and polynomial
regression is explored by building an explicit expression for the coefficients
of a polynomial regression from the weights of a given neural network, using a
Taylor expansion approach. This is achieved for single hidden layer neural
networks in regression problems. The validity of the proposed method depends on
different factors like the distribution of the synaptic potentials or the
chosen activation function. The performance of this method is empirically
tested via simulation of synthetic data generated from polynomials to train
neural networks with different structures and hyperparameters, showing that
almost identical predictions can be obtained when certain conditions are met.
Lastly, when learning from polynomial generated data, the proposed method
produces polynomials that approximate correctly the data locally.
| stat |
A Semi-Parametric Estimation Method for the Quantile Spectrum with an
Application to Earthquake Classification Using Convolutional Neural Network | In this paper, a new estimation method is introduced for the quantile
spectrum, which uses a parametric form of the autoregressive (AR) spectrum
coupled with nonparametric smoothing. The method begins with quantile
periodograms which are constructed by trigonometric quantile regression at
different quantile levels, to represent the serial dependence of time series at
various quantiles. At each quantile level, we approximate the quantile spectrum
by a function in the form of an ordinary AR spectrum. In this model, we first
compute what we call the quantile autocovariance function (QACF) by the inverse
Fourier transformation of the quantile periodogram at each quantile level.
Then, we solve the Yule-Walker equations formed by the QACF to obtain the
quantile partial autocorrelation function (QPACF) and the scale parameter.
Finally, we smooth QPACF and the scale parameter across the quantile levels
using a nonparametric smoother, convert the smoothed QPACF to AR coefficients,
and obtain the AR spectral density function. Numerical results show that the
proposed method outperforms other conventional smoothing techniques. We take
advantage of the two-dimensional property of the estimators and train a
convolutional neural network (CNN) to classify smoothed quantile periodogram of
earthquake data and achieve a higher accuracy than a similar classifier using
ordinary periodograms.
| stat |
Misplaced Confidence in Observed Power | A recently published randomized controlled trial in JAMA investigated the
impact of the selective serotonin reuptake inhibitor, escitalopram, on the risk
of major adverse events (MACE). The authors estimated a hazard ratio (HR) of
0.69 (95% CI: 0.49, 0.96; $p$ = 0.03) and then attempted to calculate how much
statistical power their study (test) had attained, and used this measure to
assess how reliable their results were. Here, we discuss why this approach,
along with other post-hoc power analyses, are highly misleading.
| stat |
Quadratic Metric Elicitation with Application to Fairness | Metric elicitation is a recent framework for eliciting performance metrics
that best reflect implicit user preferences. This framework enables a
practitioner to adjust the performance metrics based on the application,
context, and population at hand. However, available elicitation strategies have
been limited to linear (or fractional-linear) functions of predictive rates. In
this paper, we develop an approach to elicit from a wider range of complex
multiclass metrics defined by quadratic functions of rates by exploiting their
local linear structure. We apply this strategy to elicit quadratic metrics for
group-based fairness, and also discuss how it can be generalized to
higher-order polynomials. Our elicitation strategies require only relative
preference feedback and are robust to both feedback and finite sample noise.
| stat |
Lifelong Generative Modeling | Lifelong learning is the problem of learning multiple consecutive tasks in a
sequential manner, where knowledge gained from previous tasks is retained and
used to aid future learning over the lifetime of the learner. It is essential
towards the development of intelligent machines that can adapt to their
surroundings. In this work we focus on a lifelong learning approach to
unsupervised generative modeling, where we continuously incorporate newly
observed distributions into a learned model. We do so through a student-teacher
Variational Autoencoder architecture which allows us to learn and preserve all
the distributions seen so far, without the need to retain the past data nor the
past models. Through the introduction of a novel cross-model regularizer,
inspired by a Bayesian update rule, the student model leverages the information
learned by the teacher, which acts as a probabilistic knowledge store. The
regularizer reduces the effect of catastrophic interference that appears when
we learn over sequences of distributions. We validate our model's performance
on sequential variants of MNIST, FashionMNIST, PermutedMNIST, SVHN and Celeb-A
and demonstrate that our model mitigates the effects of catastrophic
interference faced by neural networks in sequential learning scenarios.
| stat |
The Geometry of Continuous Latent Space Models for Network Data | We review the class of continuous latent space (statistical) models for
network data, paying particular attention to the role of the geometry of the
latent space. In these models, the presence/absence of network dyadic ties are
assumed to be conditionally independent given the dyads? unobserved positions
in a latent space. In this way, these models provide a probabilistic framework
for embedding network nodes in a continuous space equipped with a geometry that
facilitates the description of dependence between random dyadic ties.
Specifically, these models naturally capture homophilous tendencies and triadic
clustering, among other common properties of observed networks. In addition to
reviewing the literature on continuous latent space models from a geometric
perspective, we highlight the important role the geometry of the latent space
plays on properties of networks arising from these models via intuition and
simulation. Finally, we discuss results from spectral graph theory that allow
us to explore the role of the geometry of the latent space, independent of
network size. We conclude with conjectures about how these results might be
used to infer the appropriate latent space geometry from observed networks.
| stat |
Causal Proportional Hazards Estimation with a Binary Instrumental
Variable | Instrumental variables (IV) are a useful tool for estimating causal effects
in the presence of unmeasured confounding. IV methods are well developed for
uncensored outcomes, particularly for structural linear equation models, where
simple two-stage estimation schemes are available. The extension of these
methods to survival settings is challenging, partly because of the nonlinearity
of the popular survival regression models and partly because of the
complications associated with right censoring or other survival features. We
develop a simple causal hazard ratio estimator in a proportional hazards model
with right censored data. The method exploits a special characterization of IV
which enables the use of an intuitive inverse weighting scheme that is
generally applicable to more complex survival settings with left truncation,
competing risks, or recurrent events. We rigorously establish the asymptotic
properties of the estimators, and provide plug-in variance estimators. The
proposed method can be implemented in standard software, and is evaluated
through extensive simulation studies. We apply the proposed IV method to a data
set from the Prostate, Lung, Colorectal and Ovarian cancer screening trial to
delineate the causal effect of flexible sigmoidoscopy screening on colorectal
cancer survival which may be confounded by informative noncompliance with the
assigned screening regimen.
| stat |
G-computation for continuous-time data: a comparison with inverse
probability weighting | Inverse probability weighting is increasingly used in causal inference, but
the g-computation constitutes a promising alternative. We aimed to compare the
performances of these methods with time-to-event outcomes. Given the
limitations of the interpretability of the hazard ratio for causal inference,
our target estimand of interest is the difference in restricted mean survival
times. We report the findings of an extensive simulation study showing that
both inverse probability weighting and g-computation are unbiased under a
correct model specification, but g-computation is generally more efficient. We
also analyse two real-world datasets to illustrate these results. Finally, we
update the R package RISCA to facilitate the implementation of g-computation.
| stat |
Bayesian Hierarchical Factor Regression Models to Infer Cause of Death
From Verbal Autopsy Data | In low-resource settings where vital registration of death is not routine it
is often of critical interest to determine and study the cause of death (COD)
for individuals and the cause-specific mortality fraction (CSMF) for
populations. Post-mortem autopsies, considered the gold standard for COD
assignment, are often difficult or impossible to implement due to deaths
occurring outside the hospital, expense, and/or cultural norms. For this
reason, Verbal Autopsies (VAs) are commonly conducted, consisting of a
questionnaire administered to next of kin recording demographic information,
known medical conditions, symptoms, and other factors for the decedent. This
article proposes a novel class of hierarchical factor regression models that
avoid restrictive assumptions of standard methods, allow both the mean and
covariance to vary with COD category, and can include covariate information on
the decedent, region, or events surrounding death. Taking a Bayesian approach
to inference, this work develops an MCMC algorithm and validates the FActor
Regression for Verbal Autopsy (FARVA) model in simulation experiments. An
application of FARVA to real VA data shows improved goodness-of-fit and better
predictive performance in inferring COD and CSMF over competing methods. Code
and a user manual are made available at https://github.com/kelrenmor/farva.
| stat |
Familywise Error Rate Controlling Procedures for Discrete Data | In applications such as clinical safety analysis, the data of the experiments
usually consists of frequency counts. In the analysis of such data, researchers
often face the problem of multiple testing based on discrete test statistics,
aimed at controlling family-wise error rate (FWER). Most existing FWER
controlling procedures are developed for continuous data, which are often
conservative when analyzing discrete data. By using minimal attainable
$p$-values, several FWER controlling procedures have been specifically
developed for discrete data in the literature. In this paper, by utilizing
known marginal distributions of true null $p$-values, three more powerful
stepwise procedures are developed, which are modified versions of the
conventional Bonferroni, Holm and Hochberg procedures, respectively. It is
shown that the first two procedures strongly control the FWER under arbitrary
dependence and are more powerful than the existing Tarone-type procedures,
while the last one only ensures control of the FWER in special settings.
Through extensive simulation studies, we provide numerical evidence of superior
performance of the proposed procedures in terms of the FWER control and minimal
power. A real clinical safety data is used to demonstrate applications of our
proposed procedures. An R package "MHTdiscrete" and a web application are
developed for implementing the proposed procedures.
| stat |
Robust Blocked Response-Adaptive Randomization Designs | In most clinical trials, patients are randomized with equal probability among
treatments to obtain an unbiased estimate of the treatment effect.
Response-adaptive randomization (RAR) has been proposed for ethical reasons,
where the randomization ratio is tilted successively to favor the better
performing treatment. However, the substantial disagreement regarding bias due
to time-trends in adaptive randomization is not fully recognized. The type-I
error is inflated in the traditional Bayesian RAR approaches when a time-trend
is present. In our approach, patients are assigned in blocks and the
randomization ratio is recomputed for blocks rather than traditional adaptive
randomization where it is done per patient. We further investigate the design
with a range of scenarios for both frequentist and Bayesian designs. We compare
our method with equal randomization and with different numbers of blocks
including the traditional RAR design where randomization ratio is altered
patient by patient basis. The analysis is stratified if there are two or more
patients in each block. Small blocks should be avoided due to the possibility
of not acquiring any information from the $\mu_i$. On the other hand, RAR with
large blocks has a good balance between efficiency and treating more subjects
to the better-performing treatment, while retaining blocked RAR's unique
unbiasedness.
| stat |
A Bayesian approach for the analysis of error rate studies in forensic
science | Over the past decade, the field of forensic science has received
recommendations from the National Research Council of the U.S. National Academy
of Sciences, the U.S. National Institute of Standards and Technology, and the
U.S. President's Council of Advisors on Science and Technology to study the
validity and reliability of forensic analyses. More specifically, these
committees recommend estimation of the rates of occurrence of erroneous
conclusions drawn from forensic analyses. "Black box" studies for the various
subjective feature-based comparison methods are intended for this purpose.
In general, "black box" studies often have unbalanced designs, comparisons
that are not independent, and missing data. These aspects pose difficulty in
the analysis of the results and are often ignored. Instead, interpretation of
the data relies on methods that assume independence between observations and a
balanced experiment. Furthermore, all of these projects are interpreted within
the frequentist framework and result in point estimates associated with
confidence intervals that are confusing to communicate and understand.
We propose to use an existing likelihood-free Bayesian inference method,
called Approximate Bayesian Computation (ABC), that is capable of handling
unbalanced designs, dependencies among the observations, and missing data. ABC
allows for studying the parameters of interest without recourse to incoherent
and misleading measures of uncertainty such as confidence intervals. By taking
into account information from all decision categories for a given examiner and
information from the population of examiners, our method also allows for
quantifying the risk of error for the given examiner, even when no error has
been recorded for that examiner.
We illustrate our proposed method by reanalysing the results of the "Noblis
Black Box" study by Ulery et al. in 2011.
| stat |
A General Framework for Bandit Problems Beyond Cumulative Objectives | The stochastic multi-armed bandit (MAB) problem is a common model for
sequential decision problems. In the standard setup, a decision maker has to
choose at every instant between several competing arms, each of them provides a
scalar random variable, referred to as a "reward." Nearly all research on this
topic considers the total cumulative reward as the criterion of interest. This
work focuses on other natural objectives that cannot be cast as a sum over
rewards, but rather more involved functions of the reward stream. Unlike the
case of cumulative criteria, in the problems we study here the oracle policy,
that knows the problem parameters a priori and is used to "center" the regret,
is not trivial. We provide a systematic approach to such problems, and derive
general conditions under which the oracle policy is sufficiently tractable to
facilitate the design of optimism-based (upper confidence bound) learning
policies. These conditions elucidate an interesting interplay between the arm
reward distributions and the performance metric. Our main findings are
illustrated for several commonly used objectives such as conditional
value-at-risk, mean-variance trade-offs, Sharpe-ratio, and more.
| stat |
Copulas as High-Dimensional Generative Models: Vine Copula Autoencoders | We introduce the vine copula autoencoder (VCAE), a flexible generative model
for high-dimensional distributions built in a straightforward three-step
procedure.
First, an autoencoder (AE) compresses the data into a lower dimensional
representation. Second, the multivariate distribution of the encoded data is
estimated with vine copulas. Third, a generative model is obtained by combining
the estimated distribution with the decoder part of the AE. As such, the
proposed approach can transform any already trained AE into a flexible
generative model at a low computational cost. This is an advantage over
existing generative models such as adversarial networks and variational AEs
which can be difficult to train and can impose strong assumptions on the latent
space. Experiments on MNIST, Street View House Numbers and Large-Scale
CelebFaces Attributes datasets show that VCAEs can achieve competitive results
to standard baselines.
| stat |
Bias-corrected estimator for intrinsic dimension and differential
entropy--a visual multiscale approach | Intrinsic dimension and differential entropy estimators are studied in this
paper, including their systematic bias. A pragmatic approach for joint
estimation and bias correction of these two fundamental measures is proposed.
Shared steps on both estimators are highlighted, along with their useful
consequences to data analysis. It is shown that both estimators can be
complementary parts of a single approach, and that the simultaneous estimation
of differential entropy and intrinsic dimension give meaning to each other,
where estimates at different observation scales convey different perspectives
of underlying manifolds. Experiments with synthetic and real datasets are
presented to illustrate how to extract meaning from visual inspections, and how
to compensate for biases.
| stat |
srMO-BO-3GP: A sequential regularized multi-objective constrained
Bayesian optimization for design applications | Bayesian optimization (BO) is an efficient and flexible global optimization
framework that is applicable to a very wide range of engineering applications.
To leverage the capability of the classical BO, many extensions, including
multi-objective, multi-fidelity, parallelization, latent-variable model, have
been proposed to improve the limitation of the classical BO framework. In this
work, we propose a novel multi-objective (MO) extension, called srMO-BO-3GP, to
solve the MO optimization problems in a sequential setting. Three different
Gaussian processes (GPs) are stacked together, where each of the GP is assigned
with a different task: the first GP is used to approximate the single-objective
function, the second GP is used to learn the unknown constraints, and the third
GP is used to learn the uncertain Pareto frontier. At each iteration, a MO
augmented Tchebycheff function converting MO to single-objective is adopted and
extended with a regularized ridge term, where the regularization is introduced
to smoothen the single-objective function. Finally, we couple the third GP
along with the classical BO framework to promote the richness and diversity of
the Pareto frontier by the exploitation and exploration acquisition function.
The proposed framework is demonstrated using several numerical benchmark
functions, as well as a thermomechanical finite element model for flip-chip
package design optimization.
| stat |
Efficient Inference in Multi-task Cox Process Models | We generalize the log Gaussian Cox process (LGCP) framework to model multiple
correlated point data jointly. The observations are treated as realizations of
multiple LGCPs, whose log intensities are given by linear combinations of
latent functions drawn from Gaussian process priors. The combination
coefficients are also drawn from Gaussian processes and can incorporate
additional dependencies. We derive closed-form expressions for the moments of
the intensity functions and develop an efficient variational inference
algorithm that is orders of magnitude faster than competing deterministic and
stochastic approximations of multivariate LGCP, coregionalization models, and
multi-task permanental processes. Our approach outperforms these benchmarks in
multiple problems, offering the current state of the art in modeling
multivariate point processes.
| stat |
Randomized p-values for multiple testing and their application in
replicability analysis | We are concerned with testing replicability hypotheses for many endpoints
simultaneously. This constitutes a multiple test problem with composite null
hypotheses. Traditional $p$-values, which are computed under least favourable
parameter configurations, are over-conservative in the case of composite null
hypotheses. As demonstrated in prior work, this poses severe challenges in the
multiple testing context, especially when one goal of the statistical analysis
is to estimate the proportion $\pi_0$ of true null hypotheses. Randomized
$p$-values have been proposed to remedy this issue. In the present work, we
discuss the application of randomized $p$-values in replicability analysis. In
particular, we introduce a general class of statistical models for which valid,
randomized $p$-values can be calculated easily. By means of computer
simulations, we demonstrate that their usage typically leads to a much more
accurate estimation of $\pi_0$. Finally, we apply our proposed methodology to a
real data example from genomics.
| stat |
The Implicit Regularization of Stochastic Gradient Flow for Least
Squares | We study the implicit regularization of mini-batch stochastic gradient
descent, when applied to the fundamental problem of least squares regression.
We leverage a continuous-time stochastic differential equation having the same
moments as stochastic gradient descent, which we call stochastic gradient flow.
We give a bound on the excess risk of stochastic gradient flow at time $t$,
over ridge regression with tuning parameter $\lambda = 1/t$. The bound may be
computed from explicit constants (e.g., the mini-batch size, step size, number
of iterations), revealing precisely how these quantities drive the excess risk.
Numerical examples show the bound can be small, indicating a tight relationship
between the two estimators. We give a similar result relating the coefficients
of stochastic gradient flow and ridge. These results hold under no conditions
on the data matrix $X$, and across the entire optimization path (not just at
convergence).
| stat |
Towards Clarifying the Theory of the Deconfounder | Wang and Blei (2019) studies multiple causal inference and proposes the
deconfounder algorithm. The paper discusses theoretical requirements and
presents empirical studies. Several refinements have been suggested around the
theory of the deconfounder. Among these, Imai and Jiang clarified the
assumption of "no unobserved single-cause confounders." Using their assumption,
this paper clarifies the theory. Furthermore, Ogburn et al. (2020) proposes
counterexamples to the theory. But the proposed counterexamples do not satisfy
the required assumptions.
| stat |
Reducing post-surgery recovery bed occupancy with a probabilistic
forecast model | Operations Research approaches to surgical scheduling are becoming
increasingly popular in both theory and practice. Often these models neglect
stochasticity in order to reduce the computational complexity of the problem.
We wish to provide practitioners and hospital administrative staff with a
start-of-day probabilistic forecast for the occupancy of post-surgery recovery
spaces. The model minimises the maximum expected occupancy of the recovery
unit, thus levelling the workload of hospital staff, and reducing the
likelihood of bed shortages and bottlenecks. We show that a Poisson binomial
random variable models the number of patients in the recovery when
parameterised by the surgical case sequence. A mixed integer nonlinear
programming model for the surgical case sequencing problem reduces the maximum
expected occupancy in post-surgery recovery spaces. Simulated Annealing
produces good solutions in short amounts of computational time. We evaluate the
methodology with a full year of historical data. The solution techniques reduce
maximum expected recovery occupancy by 18% on average. This alleviates a large
amount of stress on staff in the postsurgery recovery spaces, reduces the
likelihood of bottlenecks, and improves the quality of care provided to
patients.
| stat |
Climate-driven statistical models as effective predictors of local
dengue incidence in Costa Rica: A Generalized Additive Model and Random
Forest approach | Climate has been an important factor in shaping the distribution and
incidence of dengue cases in tropical and subtropical countries. In Costa Rica,
a tropical country with distinctive micro-climates, dengue has been endemic
since its introduction in 1993, inflicting substantial economic, social, and
public health repercussions. Using the number of dengue reported cases and
climate data from 2007-2017, we fitted a prediction model applying a
Generalized Additive Model (GAM) and Random Forest (RF) approach, which allowed
us to retrospectively predict the relative risk of dengue in five
climatological diverse municipalities around the country.
| stat |
On the Sample Complexity of Rank Regression from Pairwise Comparisons | We consider a rank regression setting, in which a dataset of $N$ samples with
features in $\mathbb{R}^d$ is ranked by an oracle via $M$ pairwise comparisons.
Specifically, there exists a latent total ordering of the samples; when
presented with a pair of samples, a noisy oracle identifies the one ranked
higher with respect to the underlying total ordering. A learner observes a
dataset of such comparisons and wishes to regress sample ranks from their
features. We show that to learn the model parameters with $\epsilon > 0$
accuracy, it suffices to conduct $M \in \Omega(dN\log^3 N/\epsilon^2)$
comparisons uniformly at random when $N$ is $\Omega(d/\epsilon^2)$.
| stat |
Sketching in Bayesian High Dimensional Regression With Big Data Using
Gaussian Scale Mixture Priors | Bayesian computation of high dimensional linear regression models with a
popular Gaussian scale mixture prior distribution using Markov Chain Monte
Carlo (MCMC) or its variants can be extremely slow or completely prohibitive
due to the heavy computational cost that grows in the cubic order of p, with p
as the number of features. Although a few recently developed algorithms make
the computation efficient in presence of a small to moderately large sample
size (with the complexity growing in the cubic order of n), the computation
becomes intractable when sample size n is also large. In this article we adopt
the data sketching approach to compress the n original samples by a random
linear transformation to m<<n samples in p dimensions, and compute Bayesian
regression with Gaussian scale mixture prior distributions with the randomly
compressed response vector and feature matrix. Our proposed approach yields
computational complexity growing in the cubic order of m. Another important
motivation for this compression procedure is that it anonymizes the data by
revealing little information about the original data in the course of analysis.
Our detailed empirical investigation with the Horseshoe prior from the class of
Gaussian scale mixture priors shows closely similar inference and a massive
reduction in per iteration computation time of the proposed approach compared
to the regression with the full sample. One notable contribution of this
article is to derive posterior contraction rate for high dimensional predictor
coefficient with a general class of shrinkage priors on them under data
compression/sketching. In particular, we characterize the dimension of the
compressed response vector m as a function of the sample size, number of
predictors and sparsity in the regression to guarantee accurate estimation of
predictor coefficients asymptotically, even after data compression.
| stat |
Leveraging directed causal discovery to detect latent common causes | The discovery of causal relationships is a fundamental problem in science and
medicine. In recent years, many elegant approaches to discovering causal
relationships between two variables from uncontrolled data have been proposed.
However, most of these deal only with purely directed causal relationships and
cannot detect latent common causes. Here, we devise a general method which
takes a purely directed causal discovery algorithm and modifies it so that it
can also detect latent common causes. The identifiability of the modified
algorithm depends on the identifiability of the original, as well as an
assumption that the strength of noise be relatively small. We apply our method
to two directed causal discovery algorithms, the Information Geometric Causal
Inference of (Daniusis et al., 2010) and the Kernel Conditional Deviance for
Causal Inference of (Mitrovic, Sejdinovic, and Teh, 2018), and extensively test
on synthetic data---detecting latent common causes in additive, multiplicative
and complex noise regimes---and on real data, where we are able to detect known
common causes. In addition to detecting latent common causes, our experiments
demonstrate that both modified algorithms preserve the performance of the
original directed algorithm in distinguishing directed causal relations.
| stat |
Scaling Up Bayesian Uncertainty Quantification for Inverse Problems
using Deep Neural Networks | Due to the importance of uncertainty quantification (UQ), Bayesian approach
to inverse problems has recently gained popularity in applied mathematics,
physics, and engineering. However, traditional Bayesian inference methods based
on Markov Chain Monte Carlo (MCMC) tend to be computationally intensive and
inefficient for such high dimensional problems. To address this issue, several
methods based on surrogate models have been proposed to speed up the inference
process. More specifically, the calibration-emulation-sampling (CES) scheme has
been proven to be successful in large dimensional UQ problems. In this work, we
propose a novel CES approach for Bayesian inference based on deep neural
network (DNN) models for the emulation phase. The resulting algorithm is not
only computationally more efficient, but also less sensitive to the training
set. Further, by using an Autoencoder (AE) for dimension reduction, we have
been able to speed up our Bayesian inference method up to three orders of
magnitude. Overall, our method, henceforth called \emph{Dimension-Reduced
Emulative Autoencoder Monte Carlo (DREAM)} algorithm, is able to scale Bayesian
UQ up to thousands of dimensions in physics-constrained inverse problems. Using
two low-dimensional (linear and nonlinear) inverse problems we illustrate the
validity this approach. Next, we apply our method to two high-dimensional
numerical examples (elliptic and advection-diffussion) to demonstrate its
computational advantage over existing algorithms.
| stat |
Robust and Efficient Optimization Using a Marquardt-Levenberg Algorithm
with R Package marqLevAlg | Optimization is an essential task in many computational problems. In
statistical modelling for instance, in the absence of analytical solution,
maximum likelihood estimators are often retrieved using iterative optimization
algorithms. R software already includes a variety of optimizers from
general-purpose optimization algorithms to more specific ones. Among
Newton-like methods which have good convergence properties, the
Marquardt-Levenberg algorithm (MLA) provides a particularly robust algorithm
for solving optimization problems. Newton-like methods generally have two major
limitations: (i) convergence criteria that are a little too loose, and do not
ensure convergence towards a maximum, (ii) a calculation time that is often too
long, which makes them unusable in complex problems. We propose in the
marqLevAlg package an efficient and general implementation of a modified MLA
combined with strict convergence criteria and parallel computations.
Convergence to saddle points is avoided by using the relative distance to
minimum/maximum criterion (RDM) in addition to the stability of the parameters
and of the objective function. RDM exploits the first and second derivatives to
compute the distance to a true local maximum. The independent multiple
evaluations of the objective function at each iteration used for computing
either first or second derivatives are called in parallel to allow a
theoretical speed up to the square of the number of parameters. We show through
the estimation of 7 relatively complex statistical models how parallel
implementation can largely reduce computational time. We also show through the
estimation of the same model using 3 different algorithms (BFGS of optim
routine, an E-M, and MLA) the superior efficiency of MLA to correctly and
consistently reach the maximum.
| stat |
A High Probability Analysis of Adaptive SGD with Momentum | Stochastic Gradient Descent (SGD) and its variants are the most used
algorithms in machine learning applications. In particular, SGD with adaptive
learning rates and momentum is the industry standard to train deep networks.
Despite the enormous success of these methods, our theoretical understanding of
these variants in the nonconvex setting is not complete, with most of the
results only proving convergence in expectation and with strong assumptions on
the stochastic gradients. In this paper, we present a high probability analysis
for adaptive and momentum algorithms, under weak assumptions on the function,
stochastic gradients, and learning rates. We use it to prove for the first time
the convergence of the gradients to zero in high probability in the smooth
nonconvex setting for Delayed AdaGrad with momentum.
| stat |
Proportional hazards model with partly interval censoring and its
penalized likelihood estimation | This paper considers the problem of semi-parametric proportional hazards
model fitting for interval, left and right censored survival times. We adopt a
more versatile penalized likelihood method to estimate the baseline hazard and
the regression coefficients simultaneously, where the penalty is introduced in
order to regularize the baseline hazard estimate. We present asymptotic
properties of our estimate, allowing for the possibility that it may lie on the
boundary of the parameter space. We also provide a computational method based
on marginal likelihood, which allows the regularization parameter to be
determined automatically. Comparisons of our method with other approaches are
given in simulations which demonstrate that our method has favourable
performance. A real data application involving a model for melanoma recurrence
is presented and an R package implementing the methods is available.
| stat |
Projected $t$-SNE for batch correction | Biomedical research often produces high-dimensional data confounded by batch
effects such as systematic experimental variations, different protocols and
subject identifiers. Without proper correction, low-dimensional representation
of high-dimensional data might encode and reproduce the same systematic
variations observed in the original data, and compromise the interpretation of
the results. In this article, we propose a novel procedure to remove batch
effects from low-dimensional embeddings obtained with t-SNE dimensionality
reduction. The proposed methods are based on linear algebra and constrained
optimization, leading to efficient algorithms and fast computation in many
high-dimensional settings. Results on artificial single-cell transcription
profiling data show that the proposed procedure successfully removes multiple
batch effects from t-SNE embeddings, while retaining fundamental information on
cell types. When applied to single-cell gene expression data to investigate
mouse medulloblastoma, the proposed method successfully removes batches related
with mice identifiers and the date of the experiment, while preserving clusters
of oligodendrocytes, astrocytes, and endothelial cells and microglia, which are
expected to lie in the stroma within or adjacent to the tumors.
| stat |
A Comparative Study of Imputation Methods for Multivariate Ordinal Data | Missing data remains a very common problem in large datasets, including
survey and census data containing many ordinal responses, such as political
polls and opinion surveys. Multiple imputation (MI) is usually the go-to
approach for analyzing such incomplete datasets, and there are indeed several
implementations of MI, including methods using generalized linear models,
tree-based models, and Bayesian non-parametric models. However, there is
limited research on the statistical performance of these methods for
multivariate ordinal data. In this article, we perform an empirical evaluation
of several MI methods, including MI by chained equations (MICE) using
multinomial logistic regression models, MICE using proportional odds logistic
regression models, MICE using classification and regression trees, MICE using
random forest, MI using Dirichlet process (DP) mixtures of products of
multinomial distributions, and MI using DP mixtures of multivariate normal
distributions. We evaluate the methods using simulation studies based on
ordinal variables selected from the 2018 American Community Survey (ACS). Under
our simulation settings, the results suggest that MI using proportional odds
logistic regression models, classification and regression trees and DP mixtures
of multinomial distributions generally outperform the other methods. In certain
settings, MI using multinomial logistic regression models is able to achieve
comparable performance, depending on the missing data mechanism and amount of
missing data.
| stat |
The battle of total-order sensitivity estimators | Sensitivity analysis (SA) helps identify which model inputs convey the most
uncertainty to the model output. One of the most authoritative measures in
global SA is the Sobol' total-order index ($T_i$), which can be computed with
several different estimators. Although previous comparisons exist, it is hard
to know which estimator performs best since the results are contingent on the
benchmark setting defined by the analyst (the sampling method, the distribution
of the model inputs, the number of model runs, the test function or model and
its dimensionality, the weight of higher order effects or the performance
measure selected). Here we compare several total-order estimators in an
eight-dimension hypercube where these benchmark parameters are treated as
random parameters. This arrangement significantly relaxes the dependency of the
results on the benchmark design. We observe that the most accurate estimators
are Razavi and Gupta's, Jansen's or Janon/Monod's for factor prioritization,
and Jansen's, Janon/Monod's or Azzini and Rosati's for approaching the "true"
total-order indices. The rest lag considerably behind. Our work helps analysts
navigate the myriad of total-order formulae by reducing the uncertainty in the
selection of the most appropriate estimator.
| stat |
Analysis of multiple data sequences with different distributions:
defining common principal component axes by ergodic sequence generation and
multiple reweighting composition | Principal component analysis (PCA) defines a reduced space described by PC
axes for a given multidimensional-data sequence to capture the variations of
the data. In practice, we need multiple data sequences that accurately obey
individual probability distributions and for a fair comparison of the sequences
we need PC axes that are common for the multiple sequences but properly capture
these multiple distributions. For these requirements, we present individual
ergodic samplings for these sequences and provide special reweighting for
recovering the target distributions.
| stat |
On the Theoretical Guarantees for Parameter Estimation of Gaussian
Random Field Models: A Sparse Precision Matrix Approach | Iterative methods for fitting a Gaussian Random Field (GRF) model via maximum
likelihood (ML) estimation requires solving a nonconvex optimization problem.
The problem is aggravated for anisotropic GRFs where the number of covariance
function parameters increases with the dimension. Even evaluation of the
likelihood function requires $O(n^3)$ floating point operations, where $n$
denotes the number of data locations. In this paper, we propose a new two-stage
procedure to estimate the parameters of second-order stationary GRFs. First, a
convex likelihood problem regularized with a weighted $\ell_1$-norm, utilizing
the available distance information between observation locations, is solved to
fit a sparse precision (inverse covariance) matrix to the observed data.
Second, the parameters of the covariance function are estimated by solving a
least squares problem. Theoretical error bounds for the solutions of stage I
and II problems are provided, and their tightness are investigated.
| stat |
Uncertainty in Neural Networks: Approximately Bayesian Ensembling | Understanding the uncertainty of a neural network's (NN) predictions is
essential for many purposes. The Bayesian framework provides a principled
approach to this, however applying it to NNs is challenging due to large
numbers of parameters and data. Ensembling NNs provides an easily
implementable, scalable method for uncertainty quantification, however, it has
been criticised for not being Bayesian. This work proposes one modification to
the usual process that we argue does result in approximate Bayesian inference;
regularising parameters about values drawn from a distribution which can be set
equal to the prior. A theoretical analysis of the procedure in a simplified
setting suggests the recovered posterior is centred correctly but tends to have
an underestimated marginal variance, and overestimated correlation. However,
two conditions can lead to exact recovery. We argue that these conditions are
partially present in NNs. Empirical evaluations demonstrate it has an advantage
over standard ensembling, and is competitive with variational methods.
| stat |
Regression to the Mean's Impact on the Synthetic Control Method: Bias
and Sensitivity Analysis | To make informed policy recommendations from observational data, we must be
able to discern true treatment effects from random noise and effects due to
confounding. Difference-in-Difference techniques which match treated units to
control units based on pre-treatment outcomes, such as the synthetic control
approach have been presented as principled methods to account for confounding.
However, we show that use of synthetic controls or other matching procedures
can introduce regression to the mean (RTM) bias into estimates of the average
treatment effect on the treated. Through simulations, we show RTM bias can lead
to inflated type I error rates as well as decreased power in typical policy
evaluation settings. Further, we provide a novel correction for RTM bias which
can reduce bias and attain appropriate type I error rates. This correction can
be used to perform a sensitivity analysis which determines how results may be
affected by RTM. We use our proposed correction and sensitivity analysis to
reanalyze data concerning the effects of California's Proposition 99, a
large-scale tobacco control program, on statewide smoking rates.
| stat |
The Exact Equivalence of Distance and Kernel Methods for Hypothesis
Testing | Distance-based tests, also called "energy statistics", are leading methods
for two-sample and independence tests from the statistics community.
Kernel-based tests, developed from "kernel mean embeddings", are leading
methods for two-sample and independence tests from the machine learning
community. A fixed-point transformation was previously proposed to connect the
distance methods and kernel methods for the population statistics. In this
paper, we propose a new bijective transformation between metrics and kernels.
It simplifies the fixed-point transformation, inherits similar theoretical
properties, allows distance methods to be exactly the same as kernel methods
for sample statistics and p-value, and better preserves the data structure upon
transformation. Our results further advance the understanding in distance and
kernel-based tests, streamline the code base for implementing these tests, and
enable a rich literature of distance-based and kernel-based methodologies to
directly communicate with each other.
| stat |
The Implicit Regularization of Ordinary Least Squares Ensembles | Ensemble methods that average over a collection of independent predictors
that are each limited to a subsampling of both the examples and features of the
training data command a significant presence in machine learning, such as the
ever-popular random forest, yet the nature of the subsampling effect,
particularly of the features, is not well understood. We study the case of an
ensemble of linear predictors, where each individual predictor is fit using
ordinary least squares on a random submatrix of the data matrix. We show that,
under standard Gaussianity assumptions, when the number of features selected
for each predictor is optimally tuned, the asymptotic risk of a large ensemble
is equal to the asymptotic ridge regression risk, which is known to be optimal
among linear predictors in this setting. In addition to eliciting this implicit
regularization that results from subsampling, we also connect this ensemble to
the dropout technique used in training deep (neural) networks, another strategy
that has been shown to have a ridge-like regularizing effect.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.