title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Long-range Event-level Prediction and Response Simulation for Urban
Crime and Global Terrorism with Granger Networks | Large-scale trends in urban crime and global terrorism are well-predicted by
socio-economic drivers, but focused, event-level predictions have had limited
success. Standard machine learning approaches are promising, but lack
interpretability, are generally interpolative, and ineffective for precise
future interventions with costly and wasteful false positives. Here, we are
introducing Granger Network inference as a new forecasting approach for
individual infractions with demonstrated performance far surpassing past
results, yet transparent enough to validate and extend social theory.
Considering the problem of predicting crime in the City of Chicago, we achieve
an average AUC of ~90\% for events predicted a week in advance within spatial
tiles approximately $1000$ ft across. Instead of pre-supposing that crimes
unfold across contiguous spaces akin to diffusive systems, we learn the local
transport rules from data. As our key insights, we uncover indications of
suburban bias -- how law-enforcement response is modulated by socio-economic
contexts with disproportionately negative impacts in the inner city -- and how
the dynamics of violent and property crimes co-evolve and constrain each other
-- lending quantitative support to controversial pro-active policing policies.
To demonstrate broad applicability to spatio-temporal phenomena, we analyze
terror attacks in the middle-east in the recent past, and achieve an AUC of
~80% for predictions made a week in advance, and within spatial tiles measuring
approximately 120 miles across. We conclude that while crime operates near an
equilibrium quickly dissipating perturbations, terrorism does not. Indeed
terrorism aims to destabilize social order, as shown by its dynamics being
susceptible to run-away increases in event rates under small perturbations.
| stat |
Constrained Bayesian Optimization with Max-Value Entropy Search | Bayesian optimization (BO) is a model-based approach to sequentially optimize
expensive black-box functions, such as the validation error of a deep neural
network with respect to its hyperparameters. In many real-world scenarios, the
optimization is further subject to a priori unknown constraints. For example,
training a deep network configuration may fail with an out-of-memory error when
the model is too large. In this work, we focus on a general formulation of
Gaussian process-based BO with continuous or binary constraints. We propose
constrained Max-value Entropy Search (cMES), a novel information
theoretic-based acquisition function implementing this formulation. We also
revisit the validity of the factorized approximation adopted for rapid
computation of the MES acquisition function, showing empirically that this
leads to inaccurate results. On an extensive set of real-world constrained
hyperparameter optimization problems we show that cMES compares favourably to
prior work, while being simpler to implement and faster than other constrained
extensions of Entropy Search.
| stat |
Leveraging Global Parameters for Flow-based Neural Posterior Estimation | Inferring the parameters of a stochastic model based on experimental
observations is central to the scientific method. A particularly challenging
setting is when the model is strongly indeterminate, i.e., when distinct sets
of parameters yield identical observations. This arises in many practical
situations, such as when inferring the distance and power of a radio source (is
the source close and weak or far and strong?) or when estimating the amplifier
gain and underlying brain activity of an electrophysiological experiment. In
this work, we present a method for cracking such indeterminacy by exploiting
additional information conveyed by an auxiliary set of observations sharing
global parameters. Our method extends recent developments in simulation-based
inference(SBI) based on normalizing flows to Bayesian hierarchical models. We
validate quantitatively our proposal on a motivating example amenable to
analytical solutions, and then apply it to invert a well known non-linear model
from computational neuroscience.
| stat |
More Efficient Estimation for Logistic Regression with Optimal Subsample | In this paper, we propose improved estimation method for logistic regression
based on subsamples taken according the optimal subsampling probabilities
developed in Wang et al. 2018 Both asymptotic results and numerical results
show that the new estimator has a higher estimation efficiency. We also develop
a new algorithm based on Poisson subsampling, which does not require to
approximate the optimal subsampling probabilities all at once. This is
computationally advantageous when available random-access memory is not enough
to hold the full data. Interestingly, asymptotic distributions also show that
Poisson subsampling produces a more efficient estimator if the sampling rate,
the ratio of the subsample size to the full data sample size, does not converge
to zero. We also obtain the unconditional asymptotic distribution for the
estimator based on Poisson subsampling. The proposed approach requires to use a
pilot estimator to correct biases of un-weighted estimators. We further show
that even if the pilot estimator is inconsistent, the resulting estimators are
still consistent and asymptotically normal if the model is correctly specified.
| stat |
Variational Mixture of Normalizing Flows | In the past few years, deep generative models, such as generative adversarial
networks \autocite{GAN}, variational autoencoders \autocite{vaepaper}, and
their variants, have seen wide adoption for the task of modelling complex data
distributions. In spite of the outstanding sample quality achieved by those
early methods, they model the target distributions \emph{implicitly}, in the
sense that the probability density functions induced by them are not explicitly
accessible. This fact renders those methods unfit for tasks that require, for
example, scoring new instances of data with the learned distributions.
Normalizing flows have overcome this limitation by leveraging the
change-of-variables formula for probability density functions, and by using
transformations designed to have tractable and cheaply computable Jacobians.
Although flexible, this framework lacked (until recently
\autocites{semisuplearning_nflows, RAD}) a way to introduce discrete structure
(such as the one found in mixtures) in the models it allows to construct, in an
unsupervised scenario. The present work overcomes this by using normalizing
flows as components in a mixture model and devising an end-to-end training
procedure for such a model. This procedure is based on variational inference,
and uses a variational posterior parameterized by a neural network. As will
become clear, this model naturally lends itself to (multimodal) density
estimation, semi-supervised learning, and clustering. The proposed model is
illustrated on two synthetic datasets, as well as on a real-world dataset.
Keywords: Deep generative models, normalizing flows, variational inference,
probabilistic modelling, mixture models.
| stat |
tmleCommunity: A R Package Implementing Target Maximum Likelihood
Estimation for Community-level Data | Over the past years, many applications aim to assess the causal effect of
treatments assigned at the community level, while data are still collected at
the individual level among individuals of the community. In many cases, one
wants to evaluate the effect of a stochastic intervention on the community,
where all communities in the target population receive probabilistically
assigned treatments based on a known specified mechanism (e.g., implementing a
community-level intervention policy that target stochastic changes in the
behavior of a target population of communities). The tmleCommunity package is
recently developed to implement targeted minimum loss-based estimation (TMLE)
of the effect of community-level intervention(s) at a single time point on an
individual-based outcome of interest, including the average causal effect.
Implementations of the inverse-probability-of-treatment-weighting (IPTW) and
the G-computation formula (GCOMP) are also available. The package supports
multivariate arbitrary (i.e., static, dynamic or stochastic) interventions with
a binary or continuous outcome. Besides, it allows user-specified data-adaptive
machine learning algorithms through SuperLearner, sl3 and h2oEnsemble packages.
The usage of the tmleCommunity package, along with a few examples, will be
described in this paper.
| stat |
Target-Embedding Autoencoders for Supervised Representation Learning | Autoencoder-based learning has emerged as a staple for disciplining
representations in unsupervised and semi-supervised settings. This paper
analyzes a framework for improving generalization in a purely supervised
setting, where the target space is high-dimensional. We motivate and formalize
the general framework of target-embedding autoencoders (TEA) for supervised
prediction, learning intermediate latent representations jointly optimized to
be both predictable from features as well as predictive of targets---encoding
the prior that variations in targets are driven by a compact set of underlying
factors. As our theoretical contribution, we provide a guarantee of
generalization for linear TEAs by demonstrating uniform stability, interpreting
the benefit of the auxiliary reconstruction task as a form of regularization.
As our empirical contribution, we extend validation of this approach beyond
existing static classification applications to multivariate sequence
forecasting, verifying their advantage on both linear and nonlinear recurrent
architectures---thereby underscoring the further generality of this framework
beyond feedforward instantiations.
| stat |
Online Adaptive Image Reconstruction (OnAIR) Using Dictionary Models | Sparsity and low-rank models have been popular for reconstructing images and
videos from limited or corrupted measurements. Dictionary or transform learning
methods are useful in applications such as denoising, inpainting, and medical
image reconstruction. This paper proposes a framework for online (or
time-sequential) adaptive reconstruction of dynamic image sequences from linear
(typically undersampled) measurements. We model the spatiotemporal patches of
the underlying dynamic image sequence as sparse in a dictionary, and we
simultaneously estimate the dictionary and the images sequentially from
streaming measurements. Multiple constraints on the adapted dictionary are also
considered such as a unitary matrix, or low-rank dictionary atoms that provide
additional efficiency or robustness. The proposed online algorithms are memory
efficient and involve simple updates of the dictionary atoms, sparse
coefficients, and images. Numerical experiments demonstrate the usefulness of
the proposed methods in inverse problems such as video reconstruction or
inpainting from noisy, subsampled pixels, and dynamic magnetic resonance image
reconstruction from very limited measurements.
| stat |
Automatic Forecasting using Gaussian Processes | Automatic forecasting is the task of receiving a time series and returning a
forecast for the next time steps without any human intervention. We propose an
approach for automatic forecasting based on Gaussian Processes (GPs). So far,
the main limits of GPs on this task have been the lack of a criterion for the
selection of the kernel and the long times required for training different
competing kernels. We design a fixed additive kernel, which contains the
components needed to model most time series. During training the unnecessary
components are made irrelevant by automatic relevance determination. We assign
priors to each hyperparameter. We design the priors by analyzing a separate set
of time series through a hierarchical GP. The resulting model performs very
well on different types of time series, being competitive or outperforming the
state-of-the-art approaches.Thanks to the priors, we reliably estimate the
parameters with a single restart; this speedup makes the model efficient to
train and suitable for processing a large number of time series.
| stat |
More Efficient Policy Learning via Optimal Retargeting | Policy learning can be used to extract individualized treatment regimes from
observational data in healthcare, civics, e-commerce, and beyond. One big
hurdle to policy learning is a commonplace lack of overlap in the data for
different actions, which can lead to unwieldy policy evaluation and poorly
performing learned policies. We study a solution to this problem based on
retargeting, that is, changing the population on which policies are optimized.
We first argue that at the population level, retargeting may induce little to
no bias. We then characterize the optimal reference policy and retargeting
weights in both binary-action and multi-action settings. We do this in terms of
the asymptotic efficient estimation variance of the new learning objective.
Extensive empirical results in a simulation study and a case study of
personalized job counseling demonstrate that retargeting is a fairly easy way
to significantly improve any policy learning procedure applied to observational
data.
| stat |
Bayesian Analysis of Spatial Generalized Linear Mixed Models with
Laplace Random Fields | Gaussian random field (GRF) models are widely used in spatial statistics to
capture spatially correlated error. We investigate the results of replacing
Gaussian processes with Laplace moving averages (LMAs) in spatial generalized
linear mixed models (SGLMMs). We demonstrate that LMAs offer improved
predictive power when the data exhibits localized spikes in the response.
SGLMMs with LMAs are shown to maintain analogous parameter inference and
similar computing to Gaussian SGLMMs. We propose a novel discrete space LMA
model for irregular lattices and construct conjugate samplers for LMAs with
georeferenced and areal support. We provide a Bayesian analysis of SGLMMs with
LMAs and GRFs over multiple data support and response types.
| stat |
Better lower bounds for missing species: improved non-parametric
moment-based estimation for large experiments | Estimation of the number of species or unobserved classes from a random
sample of the underlying population is a ubiquitous problem in statistics. In
classical settings, the size of the sample is usually small. New technologies
such as high-throughput DNA sequencing have allowed for the sampling of
extremely large and heterogeneous populations at scales not previously
attainable or even considered. New algorithms are required that take advantage
of the size of the data to account for heterogeneity, but are also sufficiently
fast and scale well with large data. We present a non-parametric moment-based
estimator that is both computationally efficient and is sufficiently flexible
to account for heterogeneity in the abundances of underlying population. This
estimator is based on an extension of a popular moment-based lower bound (Chao,
1984), originally developed by Harris (1959) but unattainable due to the lack
of economical algorithms to solve the system of nonlinear equation required for
estimation. We apply results from the classical moment problem to show that
solutions can be obtained efficiently, allowing for estimators that are
simultaneously conservative and use more information. This is critical for
modern genomic applications, where there may be many large experiments that
require the application of species estimation. We present applications of our
estimator to estimating T-Cell receptor repertoire and dropout in single cell
RNA-seq experiments.
| stat |
Covariate-Adjusted Inference for Differential Analysis of
High-Dimensional Networks | Differences between biological networks corresponding to disease conditions
can help delineate the underlying disease mechanisms. Existing methods for
differential network analysis do not account for dependence of networks on
covariates. As a result, these approaches may detect spurious differential
connections induced by the effect of the covariates on both the disease
condition and the network. To address this issue, we propose a general
covariate-adjusted test for differential network analysis. Our method assesses
differential network connectivity by testing the null hypothesis that the
network is the same for individuals who have identical covariates and only
differ in disease status. We show empirically in a simulation study that the
covariate-adjusted test exhibits improved type-I error control compared with
na\"ive hypothesis testing procedures that do not account for covariates. We
additionally show that there are settings in which our proposed methodology
provides improved power to detect differential connections. We illustrate our
method by applying it to detect differences in breast cancer gene co-expression
networks by subtype.
| stat |
The von Mises-Fisher Procrustes model in functional Magnetic Resonance
Imaging data | The Procrustes method allows to align matrices into a common space using
similarity transformation. However, it is an ill-posed problem, i.e., it
doesn't return a unique solution about the optimal transformation, losing
results interpretability. For that, we extend the Perturbation model, which
rephrases the Procrustes method as a statistical model, defining the von
Mises-Fisher Procrustes model. The extension is focused on specifying a
regularization term, using a proper prior distribution for the orthogonal
matrix parameter of the Perturbation model. The von Mises-Fisher distribution
is then utilized to insert prior information about the final common space
structure. Thanks to that, we resolve the no-uniqueness problem of the
Procrustes method, getting an interpretable estimator for the orthogonal matrix
transformation. Being a conjugate prior, the posterior parameter is a sort of
weighted average between the maximum likelihood and prior estimator.
Application on functional Magnetic Resonance Imaging data shows an improvement
in the group-level analysis in terms of inference and interpretability of the
results. In this case, the prior information used is the three-dimensional
voxel coordinates. It permits the construction of the location matrix parameter
of the von Mises-Fisher distribution as a similarity euclidean matrix. In this
way, we can exploit the idea that the orthogonal matrix must combine spatially
close variables, i.e., voxels. The resulting orthogonal estimators reflect the
three-dimensional structure of the voxel's space, as the final group-analysis
results.
| stat |
Recurrent machines for likelihood-free inference | Likelihood-free inference is concerned with the estimation of the parameters
of a non-differentiable stochastic simulator that best reproduce real
observations. In the absence of a likelihood function, most of the existing
inference methods optimize the simulator parameters through a handcrafted
iterative procedure that tries to make the simulated data more similar to the
observations. In this work, we explore whether meta-learning can be used in the
likelihood-free context, for learning automatically from data an iterative
optimization procedure that would solve likelihood-free inference problems. We
design a recurrent inference machine that learns a sequence of parameter
updates leading to good parameter estimates, without ever specifying some
explicit notion of divergence between the simulated data and the real data
distributions. We demonstrate our approach on toy simulators, showing promising
results both in terms of performance and robustness.
| stat |
Pitfalls and Protocols in Practice of Manufacturing Data Science | The practical application of machine learning and data science (ML/DS)
techniques present a range of procedural issues to be examined and resolve
including those relating to the data issues, methodologies, assumptions, and
applicable conditions. Each of these issues can present difficulties in
practice; particularly, associated with the manufacturing characteristics and
domain knowledge. The purpose of this paper is to highlight some of the
pitfalls that have been identified in real manufacturing application under each
of these headings and to suggest protocols to avoid the pitfalls and guide the
practical applications of the ML/DS methodologies from predictive analytics to
prescriptive analytics.
| stat |
Adding new experimental arms to randomised clinical trials: impact on
error rates | Background: Experimental treatments pass through various stages of
development. If a treatment passes through early phase experiments, the
investigators may want to assess it in a late phase randomised controlled
trial. An efficient way to do this is adding it as a new research arm to an
ongoing trial. This allows to add the new treatment while the existing arms
continue. The familywise type I error rate (FWER) is often a key quantity of
interest in any multi-arm trial. We set out to clarify how it should be
calculated when new arms are added to a trial some time after it has started.
Methods: We show how the FWER, any-pair and all-pairs powers can be
calculated when a new arm is added to a platform trial. We extend the Dunnett
probability and derive analytical formulae for the correlation between the test
statistics of the existing pairwise comparison and that of the newly added arm.
We also verify our analytical derivation via simulations.
Results: Our results indicate that the FWER depends on the shared control arm
information (i.e. individuals in continuous and binary outcomes and primary
outcome events in time-to-event outcomes) from the common control arm patients
and the allocation ratio. The FWER is driven more by the number of pairwise
comparisons and the corresponding (pairwise) Type I error rates than by the
timing of the addition of the new arms. The FWER can be estimated using
\v{S}id\'{a}k's correction if the correlation between the test statistics of
pairwise comparisons is less than 0:30.
Conclusions: The findings we present in this article can be used to design
trials with pre-planned deferred arms or to design new pairwise comparisons
within an ongoing platform trial where control of the pairwise error rate
(PWER) or FWER (for a subset of pairwise comparisons) is required.
| stat |
Breaking hypothesis testing for failure rates | We describe the utility of point processes and failure rates and the most
common point process for modeling failure rates, the Poisson point process.
Next, we describe the uniformly most powerful test for comparing the rates of
two Poisson point processes for a one-sided test (henceforth referred to as the
"rate test"). A common argument against using this test is that real world data
rarely follows the Poisson point process. We thus investigate what happens when
the distributional assumptions of tests like these are violated and the test
still applied. We find a non-pathological example (using the rate test on a
Compound Poisson distribution with Binomial compounding) where violating the
distributional assumptions of the rate test make it perform better (lower error
rates). We also find that if we replace the distribution of the test statistic
under the null hypothesis with any other arbitrary distribution, the
performance of the test (described in terms of the false negative rate to false
positive rate trade-off) remains exactly the same. Next, we compare the
performance of the rate test to a version of the Wald test customized to the
Negative Binomial point process and find it to perform very similarly while
being much more general and versatile. Finally, we discuss the applications to
Microsoft Azure. The code for all experiments performed is open source and
linked in the introduction.
| stat |
COT-GAN: Generating Sequential Data via Causal Optimal Transport | We introduce COT-GAN, an adversarial algorithm to train implicit generative
models optimized for producing sequential data. The loss function of this
algorithm is formulated using ideas from Causal Optimal Transport (COT), which
combines classic optimal transport methods with an additional temporal
causality constraint. Remarkably, we find that this causality condition
provides a natural framework to parameterize the cost function that is learned
by the discriminator as a robust (worst-case) distance, and an ideal mechanism
for learning time dependent data distributions. Following Genevay et al.\
(2018), we also include an entropic penalization term which allows for the use
of the Sinkhorn algorithm when computing the optimal transport cost. Our
experiments show effectiveness and stability of COT-GAN when generating both
low- and high-dimensional time series data. The success of the algorithm also
relies on a new, improved version of the Sinkhorn divergence which demonstrates
less bias in learning.
| stat |
Performance and Application of Estimators for the Value of an Optimal
Dynamic Treatment Rule | Given an (optimal) dynamic treatment rule, it may be of interest to evaluate
that rule -- that is, to ask the causal question: what is the expected outcome
had every subject received treatment according to that rule? In this paper, we
study the performance of estimators that approximate the true value of: 1) an
$a$ $priori$ known dynamic treatment rule 2) the true, unknown optimal dynamic
treatment rule (ODTR); 3) an estimated ODTR, a so-called "data-adaptive
parameter," whose true value depends on the sample. Using simulations of
point-treatment data, we specifically investigate: 1) the impact of
increasingly data-adaptive estimation of nuisance parameters and/or of the ODTR
on performance; 2) the potential for improved efficiency and bias reduction
through the use of semiparametric efficient estimators; and, 3) the importance
of sample splitting based on CV-TMLE for accurate inference. In the simulations
considered, there was very little cost and many benefits to using the
cross-validated targeted maximum likelihood estimator (CV-TMLE) to estimate the
value of the true and estimated ODTR; importantly, and in contrast to non
cross-validated estimators, the performance of CV-TMLE was maintained even when
highly data-adaptive algorithms were used to estimate both nuisance parameters
and the ODTR. In addition, we apply these estimators for the value of the rule
to the "Interventions" Study, an ongoing randomized controlled trial, to
identify whether assigning cognitive behavioral therapy (CBT) to criminal
justice-involved adults with mental illness using an ODTR significantly reduces
the probability of recidivism, compared to assigning CBT in a
non-individualized way.
| stat |
What You See May Not Be What You Get: UCB Bandit Algorithms Robust to
{\epsilon}-Contamination | Motivated by applications of bandit algorithms in education, we consider a
stochastic multi-armed bandit problem with $\varepsilon$-contaminated rewards.
We allow an adversary to give arbitrary unbounded contaminated rewards with
full knowledge of the past and future. We impose the constraint that for each
time $t$ the proportion of contaminated rewards for any action is less than or
equal to $\varepsilon$. We derive concentration inequalities for two robust
mean estimators for sub-Gaussian distributions in the
$\varepsilon$-contamination context. We define the $\varepsilon$-contaminated
stochastic bandit problem and use our robust mean estimators to give two
variants of a robust Upper Confidence Bound (UCB) algorithm, crUCB. Using
regret derived from only the underlying stochastic rewards, both variants of
crUCB achieve $\mathcal{O} (\sqrt{KT\log T})$ regret for small enough
contamination proportions. Our simulations assume small horizons, reflecting
the newly explored setting of bandits in education. We show that in certain
adversarial regimes crUCB not only outperforms algorithms designed for
stochastic (UCB1) and adversarial (EXP3) bandits but also those that have "best
of both worlds" guarantees (EXP3++ and TsallisInf) even when our constraint on
the proportion of contaminated rewards is broken.
| stat |
Basis Expansions for Functional Snippets | Estimation of mean and covariance functions is fundamental for functional
data analysis. While this topic has been studied extensively in the literature,
a key assumption is that there are enough data in the domain of interest to
estimate both the mean and covariance functions. In this paper, we investigate
mean and covariance estimation for functional snippets in which observations
from a subject are available only in an interval of length strictly (and often
much) shorter than the length of the whole interval of interest. For such a
sampling plan, no data is available for direct estimation of the off-diagonal
region of the covariance function. We tackle this challenge via a basis
representation of the covariance function. The proposed approach allows one to
consistently estimate an infinite-rank covariance function from functional
snippets. We establish the convergence rates for the proposed estimators and
illustrate their finite-sample performance via simulation studies and two data
applications.
| stat |
Bias in Zipf's Law Estimators | The prevailing maximum likelihood estimators for inferring power law models
from rank-frequency data are biased. The source of this bias is an
inappropriate likelihood function. The correct likelihood function is derived
and shown to be computationally intractable. A more computationally efficient
method of approximate Bayesian computation (ABC) is explored. This method is
shown to have less bias for data generated from idealised rank-frequency
Zipfian distributions. However, the existing estimators and the ABC estimator
described here assume that words are drawn from a simple probability
distribution, while language is a much more complex process. We show that this
false assumption leads to continued biases when applying any of these methods
to natural language to estimate Zipf exponents. We recommend caution when
applying maximum likelihood estimation to investigate power laws in
rank-frequency data, and suggest that researchers instead consider using
graphical methods such as ordinary least squares and/or transform the data to a
frequency-size representation.
| stat |
A joint model for multiple dynamic processes and clinical endpoints:
application to Alzheimer's disease | As other neurodegenerative diseases, Alzheimer's disease, the most frequent
dementia in the elderly, is characterized by multiple progressive impairments
in the brain structure and in clinical functions such as cognitive functioning
and functional disability. Until recently, these components were mostly studied
independently since no joint model for multivariate longitudinal data and time
to event was available in the statistical community. Yet, these components are
fundamentally inter-related in the degradation process towards dementia and
should be analyzed together. We thus propose a joint model to simultaneously
describe the dynamics of multiple correlated components. Each component,
defined as a latent process, is measured by one or several continuous markers
(not necessarily Gaussian). Rather than considering the associated time to
diagnosis as in standard joint models, we assume diagnosis corresponds to the
passing above a covariate-specific threshold (to be estimated) of a
pathological process which is modelled as a combination of the
component-specific latent processes. This definition captures the clinical
complexity of diagnoses such as dementia diagnosis but also benefits from
simplifications for the computation of Maximum Likelihood Estimates. We show
that the model and estimation procedure can also handle competing clinical
endpoints. The estimation procedure, implemented in a R package, is validated
by simulations and the method is illustrated on a large French population-based
cohort of cerebral aging in which we focused on the dynamics of three clinical
manifestations and the associated risk of dementia and death before dementia.
| stat |
Moments of the multivariate Beta distribution | In this paper, we extend Beta distribution to 2 by 2 matrix and give the
analytical formula for its moments. Our analytical formula can be used to
analyze the asymptotic behavior of Beta distribution for 2 by 2 matrix.
| stat |
Gaussian Process Learning via Fisher Scoring of Vecchia's Approximation | We derive a single pass algorithm for computing the gradient and Fisher
information of Vecchia's Gaussian process loglikelihood approximation, which
provides a computationally efficient means for applying the Fisher scoring
algorithm for maximizing the loglikelihood. The advantages of the optimization
techniques are demonstrated in numerical examples and in an application to Argo
ocean temperature data. The new methods are more accurate and much faster than
an optimization method that uses only function evaluations, especially when the
covariance function has many parameters. This allows practitioners to fit
nonstationary models to large spatial and spatial-temporal datasets.
| stat |
On the Estimation of Entropy in the FastICA Algorithm | The fastICA method is a popular dimension reduction technique used to reveal
patterns in data. Here we show both theoretically and in practice that the
approximations used in fastICA can result in patterns not being successfully
recognised. We demonstrate this problem using a two-dimensional example where a
clear structure is immediately visible to the naked eye, but where the
projection chosen by fastICA fails to reveal this structure. This implies that
care is needed when applying fastICA. We discuss how the problem arises and how
it is intrinsically connected to the approximations that form the basis of the
computational efficiency of fastICA.
| stat |
A streaming algorithm for bivariate empirical copulas | Empirical copula functions can be used to model the dependence structure of
multivariate data. The Greenwald and Khanna algorithm is adapted in order to
provide a space-memory efficient approximation to the empirical copula function
of a bivariate stream of data. A succinct space-memory efficient summary of
values seen in the stream up to a certain time is maintained and can be queried
at any point to return an approximation to the empirical bivariate copula
function with guaranteed error bounds. An example then illustrates how these
summaries can be used as a tool to compute approximations to higher dimensional
copula decompositions containing bivariate copulas. The computational benefits
and approximation error of the algorithm is theoretically and numerically
assessed.
| stat |
Every Local Minimum Value is the Global Minimum Value of Induced Model
in Non-convex Machine Learning | For nonconvex optimization in machine learning, this article proves that
every local minimum achieves the globally optimal value of the perturbable
gradient basis model at any differentiable point. As a result, nonconvex
machine learning is theoretically as supported as convex machine learning with
a handcrafted basis in terms of the loss at differentiable local minima, except
in the case when a preference is given to the handcrafted basis over the
perturbable gradient basis. The proofs of these results are derived under mild
assumptions. Accordingly, the proven results are directly applicable to many
machine learning models, including practical deep neural networks, without any
modification of practical methods. Furthermore, as special cases of our general
results, this article improves or complements several state-of-the-art
theoretical results on deep neural networks, deep residual networks, and
overparameterized deep neural networks with a unified proof technique and novel
geometric insights. A special case of our results also contributes to the
theoretical foundation of representation learning.
| stat |
Infinite attention: NNGP and NTK for deep attention networks | There is a growing amount of literature on the relationship between wide
neural networks (NNs) and Gaussian processes (GPs), identifying an equivalence
between the two for a variety of NN architectures. This equivalence enables,
for instance, accurate approximation of the behaviour of wide Bayesian NNs
without MCMC or variational approximations, or characterisation of the
distribution of randomly initialised wide NNs optimised by gradient descent
without ever running an optimiser. We provide a rigorous extension of these
results to NNs involving attention layers, showing that unlike single-head
attention, which induces non-Gaussian behaviour, multi-head attention
architectures behave as GPs as the number of heads tends to infinity. We
further discuss the effects of positional encodings and layer normalisation,
and propose modifications of the attention mechanism which lead to improved
results for both finite and infinitely wide NNs. We evaluate attention kernels
empirically, leading to a moderate improvement upon the previous
state-of-the-art on CIFAR-10 for GPs without trainable kernels and advanced
data preprocessing. Finally, we introduce new features to the Neural Tangents
library (Novak et al., 2020) allowing applications of NNGP/NTK models, with and
without attention, to variable-length sequences, with an example on the IMDb
reviews dataset.
| stat |
Investigation of finite-sample properties of robust location and scale
estimators | When the experimental data set is contaminated, we usually employ robust
alternatives to common location and scale estimators such as the sample median
and Hodges-Lehmann estimators for location and the sample median absolute
deviation and Shamos estimators for scale. It is well known that these
estimators have high positive asymptotic breakdown points and are
Fisher-consistent as the sample size tends to infinity. To the best of our
knowledge, the finite-sample properties of these estimators, depending on the
sample size, have not well been studied in the literature. In this paper, we
fill this gap by providing their closed-form finite-sample breakdown points and
calculating the unbiasing factors and relative efficiencies of the robust
estimators through the extensive Monte Carlo simulations up to the sample size
100. The numerical study shows that the unbiasing factor improves the
finite-sample performance significantly. In addition, we provide the predicted
values for the unbiasing factors obtained by using the least squares method
which can be used for the case of sample size more than 100.
| stat |
A Latent Survival Analysis Enabled Simulation Platform For Nursing Home
Staffing Strategy Evaluation | Nursing homes are critical facilities for caring frail older adults with
round-the-clock formal care and personal assistance. To ensure quality care for
nursing home residents, adequate staffing level is of great importance. Current
nursing home staffing practice is mainly based on experience and regulation.
The objective of this paper is to investigate the viability of experience-based
and regulation-based strategies, as well as alternative staffing strategies to
minimize labor costs subject to heterogeneous service demand of nursing home
residents under various scenarios of census. We propose a data-driven analysis
framework to model heterogeneous service demand of nursing home residents and
further identify appropriate staffing strategies by combing survival model and
computer simulation techniques as well as domain knowledge. Specifically, in
the analysis, we develop an agent-based simulation tool consisting of four main
modules, namely individual length of stay predictor, individual daily staff
time generator, facility level staffing strategy evaluator, and graphical user
interface. We use real nursing home data to validate the proposed model, and
demonstrate that the identified staffing strategy significantly reduces the
total labor cost of certified nursing assistants compared to the benchmark
strategies. Additionally, the proposed length of stay predictive model that
considers multiple discharge dispositions exhibits superior accuracy and offers
better staffing decisions than those without the consideration. Further, we
construct different census scenarios of nursing home residents to demonstrate
the capability of the proposed framework in helping adjust staffing decisions
of nursing home administrators in various realistic settings.
| stat |
Outlier detection in non-elliptical data by kernel MRCD | The minimum regularized covariance determinant method (MRCD) is a robust
estimator for multivariate location and scatter, which detects outliers by
fitting a robust covariance matrix to the data. Its regularization ensures that
the covariance matrix is well-conditioned in any dimension. The MRCD assumes
that the non-outlying observations are roughly elliptically distributed, but
many datasets are not of that form. Moreover, the computation time of MRCD
increases substantially when the number of variables goes up, and nowadays
datasets with many variables are common. The proposed Kernel Minimum
Regularized Covariance Determinant (KMRCD) estimator addresses both issues. It
is not restricted to elliptical data because it implicitly computes the MRCD
estimates in a kernel induced feature space. A fast algorithm is constructed
that starts from kernel-based initial estimates and exploits the kernel trick
to speed up the subsequent computations. Based on the KMRCD estimates, a rule
is proposed to flag outliers. The KMRCD algorithm performs well in simulations,
and is illustrated on real-life data.
| stat |
Partition-Mallows Model and Its Inference for Rank Aggregation | Learning how to aggregate ranking lists has been an active research area for
many years and its advances have played a vital role in many applications
ranging from bioinformatics to internet commerce. The problem of discerning
reliability of rankers based only on the rank data is of great interest to many
practitioners, but has received less attention from researchers. By dividing
the ranked entities into two disjoint groups, i.e., relevant and
irrelevant/background ones, and incorporating the Mallows model for the
relative ranking of relevant entities, we propose a framework for rank
aggregation that can not only distinguish quality differences among the rankers
but also provide the detailed ranking information for relevant entities.
Theoretical properties of the proposed approach are established, and its
advantages over existing approaches are demonstrated via simulation studies and
real-data applications. Extensions of the proposed method to handle partial
ranking lists and conduct covariate-assisted rank aggregation are also
discussed.
| stat |
TCA and TLRA: A comparison on contingency tables and compositional data | There are two popular general approaches for the analysis and visualization
of a contingency table and a compositional data set: Correspondence analysis
(CA) and log ratio analysis (LRA). LRA includes two independently well
developed methods: association models and compositional data analysis. The
application of either CA or LRA to a contingency table or to compositional data
set includes a preprocessing centering step. In CA the centering step is
multiplicative, while in LRA it is log bi-additive. A preprocessed matrix is
double-centered, so it is a residuel matrix; which implies that it affects the
final results of the analysis. This paper introduces a novel index named the
intrinsic measure of the quality of the signs of the residuals (QSR) for the
choice of the preprocessing, and consequently of the method. The criterion is
based on taxicab singular value decomposition (TSVD) on which the package
TaxicabCA in R is developed. We present a minimal R script that can be executed
to obtain the numerical results and the maps in this paper. Three relatively
small sized data sets available freely on the web are used as examples.
| stat |
Kernel density estimation for directional-linear data | A nonparametric kernel density estimator for directional-linear data is
introduced. The proposal is based on a product kernel accounting for the
different nature of both (directional and linear) components of the random
vector. Expressions for bias, variance and Mean Integrated Squared Error (MISE)
are derived, jointly with an asymptotic normality result for the proposed
estimator. For some particular distributions, an explicit formula for the MISE
is obtained and compared with its asymptotic version, both for directional and
directional-linear kernel density estimators. In this same setting a closed
expression for the bootstrap MISE is also derived.
| stat |
Improving the replicability of results from a single psychological
experiment | We identify two aspects of selective inference as major obstacles for
replicability. The practice of highlighting a subset of statistical results
without taking into consideration the multiple comparisons made in the analysis
from which they were selected. The file-drawer effect, the tendency to only
publish statistically significant results. We propose to address the first
issue by controlling the FDR using the hierarchical Benjamini-Hochberg
procedure of Benjamini and Bogomolov. To address the second issue, we propose
constructing confidence intervals and estimators that are conditioned on
passing a threshold level of statistical significance. We apply our proposed
methodologies to the 100 experimental psychology studies for which replication
was tested as part of the Reproducibility Project in Psychology (RPP). We
showed that these two simple-to-use tools can enhance the replicability of
published findings without sacrificing statistical power, and are essential
even when adhering to alternative methods proposed for addressing the
replicability crisis in psychology.
| stat |
On Negative Transfer and Structure of Latent Functions in Multi-output
Gaussian Processes | The multi-output Gaussian process ($\mathcal{MGP}$) is based on the
assumption that outputs share commonalities, however, if this assumption does
not hold negative transfer will lead to decreased performance relative to
learning outputs independently or in subsets. In this article, we first define
negative transfer in the context of an $\mathcal{MGP}$ and then derive
necessary conditions for an $\mathcal{MGP}$ model to avoid negative transfer.
Specifically, under the convolution construction, we show that avoiding
negative transfer is mainly dependent on having a sufficient number of latent
functions $Q$ regardless of the flexibility of the kernel or inference
procedure used. However, a slight increase in $Q$ leads to a large increase in
the number of parameters to be estimated. To this end, we propose two latent
structures that scale to arbitrarily large datasets, can avoid negative
transfer and allow any kernel or sparse approximations to be used within. These
structures also allow regularization which can provide consistent and automatic
selection of related outputs.
| stat |
How Ominous is the Future Global Warming Premonition? | Global warming, the phenomenon of increasing global average temperature in
the recent decades, is receiving wide attention due to its very significant
adverse effects on climate. Whether global warming will continue even in the
future, is a question that is most important to investigate. In this regard,
the so-called general circulation models (GCMs) have attempted to project the
future climate, and nearly all of them exhibit alarming rates of global
temperature rise in the future.
Although global warming in the current time frame is undeniable, it is
important to assess the validity of the future predictions of the GCMs. In this
article, we attempt such a study using our recently-developed Bayesian multiple
testing paradigm for model selection in inverse regression problems. The model
we assume for the global temperature time series is based on Gaussian process
emulation of the black box scenario, realistically treating the dynamic
evolution of the time series as unknown.
We apply our ideas to datasets available from the Intergovernmental Panel on
Climate Change (IPCC) website. The best GCM models selected by our method under
different assumptions on future climate change scenarios do not convincingly
support the present global warming pattern when only the future predictions are
considered known. Using our Gaussian process idea, we also forecast the future
temperature time series given the current one. Interestingly, our results do
not support drastic future global warming predicted by almost all the GCM
models.
| stat |
Graph Enhanced High Dimensional Kernel Regression | In this paper, the flexibility, versatility and predictive power of kernel
regression are combined with now lavishly available network data to create
regression models with even greater predictive performances. Building from
previous work featuring generalized linear models built in the presence of
network cohesion data, we construct a kernelized extension that captures
subtler nonlinearities in extremely high dimensional spaces and also produces
far better predictive performances. Applications of seamless yet substantial
adaptation to simulated and real-life data demonstrate the appeal and strength
of our work.
| stat |
Properties of restricted randomization with implications for
experimental design | Recently, there as been an increasing interest in the use of heavily
restricted randomization designs which enforces balance on observed covariates
in randomized controlled trials. However, when restrictions are too strict,
there is a risk that the treatment effect estimator will have a very high mean
squared error. In this paper, we formalize this risk and propose a novel
combinatoric-based approach to describe and address this issue. First, some
known properties of complete randomization and restricted randomization are
re-proven using basic combinatorics. Second, a novel diagnostic measure that
only use the information embedded in the combinatorics of the design is
proposed. Finally, we identify situations in which restricted designs can lead
to an increased risk of getting a high mean squared error and discuss how our
diagnostic measure can be used to detect and avoid such designs. Our results
have implications for any restricted randomization design and can be used to
evaluate the trade-off between enforcing balance on observed covariates and
avoiding too restrictive designs.
| stat |
Sharpening the Rosenbaum Sensitivity Bounds to Address Concerns About
Interactions Between Observed and Unobserved Covariates | In observational studies, it is typically unrealistic to assume that
treatments are randomly assigned, even conditional on adjusting for all
observed covariates. Therefore, a sensitivity analysis is often needed to
examine how hidden biases due to unobserved covariates would affect inferences
on treatment effects. In matched observational studies where each treated unit
is matched to one or multiple untreated controls for observed covariates, the
Rosenbaum bounds sensitivity analysis is one of the most popular sensitivity
analysis models. In this paper, we show that in the presence of interactions
between observed and unobserved covariates, directly applying the Rosenbaum
bounds will almost inevitably exaggerate the report of sensitivity of causal
conclusions to hidden bias. We give sharper odds ratio bounds to fix this
deficiency. We illustrate our new method through studying the effect of
anger/hostility tendency on the risk of having heart problems.
| stat |
Fractional Underdamped Langevin Dynamics: Retargeting SGD with Momentum
under Heavy-Tailed Gradient Noise | Stochastic gradient descent with momentum (SGDm) is one of the most popular
optimization algorithms in deep learning. While there is a rich theory of SGDm
for convex problems, the theory is considerably less developed in the context
of deep learning where the problem is non-convex and the gradient noise might
exhibit a heavy-tailed behavior, as empirically observed in recent studies. In
this study, we consider a \emph{continuous-time} variant of SGDm, known as the
underdamped Langevin dynamics (ULD), and investigate its asymptotic properties
under heavy-tailed perturbations. Supported by recent studies from statistical
physics, we argue both theoretically and empirically that the heavy-tails of
such perturbations can result in a bias even when the step-size is small, in
the sense that \emph{the optima of stationary distribution} of the dynamics
might not match \emph{the optima of the cost function to be optimized}. As a
remedy, we develop a novel framework, which we coin as \emph{fractional} ULD
(FULD), and prove that FULD targets the so-called Gibbs distribution, whose
optima exactly match the optima of the original cost. We observe that the Euler
discretization of FULD has noteworthy algorithmic similarities with
\emph{natural gradient} methods and \emph{gradient clipping}, bringing a new
perspective on understanding their role in deep learning. We support our theory
with experiments conducted on a synthetic model and neural networks.
| stat |
Establishing phone-pair co-usage by comparing mobility patterns | In forensic investigations it is often of value to establish whether two
phones were used by the same person during a given time period. We present a
method that uses time and location of cell tower registrations of mobile phones
to assess the strength of evidence that any pair of phones were used by the
same person. The method is transparent as it uses logistic regression to
discriminate between the hypotheses of same and different user, and a standard
kernel density estimation to quantify the weight of evidence in terms of a
likelihood ratio. We further add to previous theoretical work by training and
validating our method on real world data, paving the way for application in
practice. The method shows good performance under different modeling choices
and robustness under lower quantity or quality of data. We discuss practical
usage in court.
| stat |
Spatiotemporal Analysis of Ridesourcing and Taxi Demand by Taxi zones in
New York City | The burst of demand for TNCs has significantly changed the transportation
landscape and dramatically disrupted the Vehicle for Hire (VFH) market that
used to be dominated by taxicabs for many years. Since first being introduced
by Uber in 2009, ridesourcing companies have rapidly penetrated the market.
This paper aims to investigate temporal and spatial patterns in taxi and TNC
usage based on data at the taxi zone level in New York City. Specifically, we
fit suitable time series models to estimate the temporal patterns. Next, we
filter out the temporal effects and investigate spatial dependence in the
residuals using global and local Moran's I statistics. We discuss the
relationship between the spatial correlations and the demographic and land use
effects at the taxi zone level. Estimating and removing these effects via a
multiple linear regression (MLR) model and recomputing the Moran's I statistics
on the resulting residuals enables us to investigate spatial dependence after
accounting for these effects. Our analysis indicates interesting patterns in
spatial correlations between taxi zones in NYC and over time, indicating that
predictive modeling of ridesourcing usage must incorporate both temporal and
spatial dependence.
| stat |
Kernel quadrature with DPPs | We study quadrature rules for functions from an RKHS, using nodes sampled
from a determinantal point process (DPP). DPPs are parametrized by a kernel,
and we use a truncated and saturated version of the RKHS kernel. This link
between the two kernels, along with DPP machinery, leads to relatively tight
bounds on the quadrature error, that depends on the spectrum of the RKHS
kernel. Finally, we experimentally compare DPPs to existing kernel-based
quadratures such as herding, Bayesian quadrature, or leverage score sampling.
Numerical results confirm the interest of DPPs, and even suggest faster rates
than our bounds in particular cases.
| stat |
Scalable Approximate Bayesian Computation for Growing Network Models via
Extrapolated and Sampled Summaries | Approximate Bayesian computation (ABC) is a simulation-based likelihood-free
method applicable to both model selection and parameter estimation. ABC
parameter estimation requires the ability to forward simulate datasets from a
candidate model, but because the sizes of the observed and simulated datasets
usually need to match, this can be computationally expensive. Additionally,
since ABC inference is based on comparisons of summary statistics computed on
the observed and simulated data, using computationally expensive summary
statistics can lead to further losses in efficiency. ABC has recently been
applied to the family of mechanistic network models, an area that has
traditionally lacked tools for inference and model choice. Mechanistic models
of network growth repeatedly add nodes to a network until it reaches the size
of the observed network, which may be of the order of millions of nodes. With
ABC, this process can quickly become computationally prohibitive due to the
resource intensive nature of network simulations and evaluation of summary
statistics. We propose two methodological developments to enable the use of ABC
for inference in models for large growing networks. First, to save time needed
for forward simulating model realizations, we propose a procedure to
extrapolate (via both least squares and Gaussian processes) summary statistics
from small to large networks. Second, to reduce computation time for evaluating
summary statistics, we use sample-based rather than census-based summary
statistics. We show that the ABC posterior obtained through this approach,
which adds two additional layers of approximation to the standard ABC, is
similar to a classic ABC posterior. Although we deal with growing network
models, both extrapolated summaries and sampled summaries are expected to be
relevant in other ABC settings where the data are generated incrementally.
| stat |
Computer model calibration based on image warping metrics: an
application for sea ice deformation | Arctic sea ice plays an important role in the global climate. Sea ice models
governed by physical equations have been used to simulate the state of the ice
including characteristics such as ice thickness, concentration, and motion.
More recent models also attempt to capture features such as fractures or leads
in the ice. These simulated features can be partially misaligned or misshapen
when compared to observational data, whether due to numerical approximation or
incomplete physics. In order to make realistic forecasts and improve
understanding of the underlying processes, it is necessary to calibrate the
numerical model to field data. Traditional calibration methods based on
generalized least-square metrics are flawed for linear features such as sea ice
cracks. We develop a statistical emulation and calibration framework that
accounts for feature misalignment and misshapenness, which involves optimally
aligning model output with observed features using cutting edge image
registration techniques. This work can also have application to other physical
models which produce coherent structures.
| stat |
Optimal allocation of subjects in a cluster randomized trial with fixed
number of clusters when the ICCs or costs are heterogeneous over clusters | The intra-cluster correlation coefficient (ICC) plays an important role while
designing the cluster randomized trials (CRTs). Often optimal CRTs are designed
assuming that the magnitude of the ICC is constant across the clusters.
However, this assumption is hardly satisfied. In some applications, the precise
information about the cluster specific correlation is known in advance. In this
article, we propose an optimal design with non-constant ICC across the
clusters. Also in many situations, the cost of sampling of an observation from
a particular cluster may differ from that of some other cluster. An optimal
design in those scenarios is also obtained assuming unequal costs of sampling
from different clusters. The theoretical findings are supplemented by thorough
numerical examples.
| stat |
Non-separable Models with High-dimensional Data | This paper studies non-separable models with a continuous treatment when the
dimension of the control variables is high and potentially larger than the
effective sample size. We propose a three-step estimation procedure to estimate
the average, quantile, and marginal treatment effects. In the first stage we
estimate the conditional mean, distribution, and density objects by penalized
local least squares, penalized local maximum likelihood estimation, and
numerical differentiation, respectively, where control variables are selected
via a localized method of L1-penalization at each value of the continuous
treatment. In the second stage we estimate the average and marginal
distribution of the potential outcome via the plug-in principle. In the third
stage, we estimate the quantile and marginal treatment effects by inverting the
estimated distribution function and using the local linear regression,
respectively. We study the asymptotic properties of these estimators and
propose a weighted-bootstrap method for inference. Using simulated and real
datasets, we demonstrate that the proposed estimators perform well in finite
samples.
| stat |
Infinite-channel deep stable convolutional neural networks | The interplay between infinite-width neural networks (NNs) and classes of
Gaussian processes (GPs) is well known since the seminal work of Neal (1996).
While numerous theoretical refinements have been proposed in the recent years,
the interplay between NNs and GPs relies on two critical distributional
assumptions on the NN's parameters: A1) finite variance; A2) independent and
identical distribution (iid). In this paper, we consider the problem of
removing A1 in the general context of deep feed-forward convolutional NNs. In
particular, we assume iid parameters distributed according to a stable
distribution and we show that the infinite-channel limit of a deep feed-forward
convolutional NNs, under suitable scaling, is a stochastic process with
multivariate stable finite-dimensional distributions. Such a limiting
distribution is then characterized through an explicit backward recursion for
its parameters over the layers. Our contribution extends results of Favaro et
al. (2020) to convolutional architectures, and it paves the way to expand
exciting recent lines of research that rely on classes of GP limits.
| stat |
A response-matrix-centred approach to presenting cross-section
measurements | The current canonical approach to publishing cross-section data is to unfold
the reconstructed distributions. Detector effects like efficiency and smearing
are undone mathematically, yielding distributions in true event properties.
This is an ill-posed problem, as even small statistical variations in the
reconstructed data can lead to large changes in the unfolded spectra.
This work presents an alternative or complementary approach: the
response-matrix-centred forward-folding approach. It offers a convenient way to
forward-fold model expectations in truth space to reconstructed quantities.
These can then be compared to the data directly, similar to what is usually
done with full detector simulations within the experimental collaborations. For
this, the detector response (efficiency and smearing) is parametrised as a
matrix. The effects of the detector on the measurement of a given model is
simulated by simply multiplying the binned truth expectation values by this
response matrix.
Systematic uncertainties in the detector response are handled by providing a
set of matrices according to the prior distribution of the detector properties
and marginalising over them. Background events can be included in the
likelihood calculation by giving background events their own bins in truth
space.
To facilitate a straight-forward use of response matrices, a new software
framework has been developed: the Response Matrix Utilities (ReMU). ReMU is a
Python package distributed via the Python Package Index. It only uses widely
available, standard scientific Python libraries and does not depend on any
custom experiment-specific software. It offers all methods needed to build
response matrices from Monte Carlo data sets, use the response matrix to
forward-fold truth-level model predictions, and compare the predictions to real
data using Bayesian or frequentist statistical inference.
| stat |
Economic variable selection | Regression plays a key role in many research areas and its variable selection
is a classic and major problem. This study emphasizes cost of predictors to be
purchased for future use, when we select a subset of them. Its economic aspect
is naturally formalized by the decision-theoretic approach. In addition, two
Bayesian approaches are proposed to address uncertainty about model parameters
and models: the restricted and extended approaches, which lead us to rethink
about model averaging. From objective, rule-based, or robust Bayes point of
view, the former is preferred. Proposed method is applied to three popular
datasets for illustration.
| stat |
Bayesian experimental design without posterior calculations: an
adversarial approach | Most computational approaches to Bayesian experimental design require making
posterior calculations, such evidence estimates, repeatedly for a large number
of potential designs and/or simulated datasets. This can be expensive and
prohibit scaling up these methods to models with many parameters, or designs
with many unknowns to select. We introduce an efficient alternative approach
without posterior calculations, based on optimising the expected trace of the
Fisher information, as discussed by Walker (2016). We illustrate drawbacks of
this approach, including lack of invariance to reparameterisation and
encouraging designs which are informative about one parameter combination but
not any others. We show these can be avoided by using an adversarial approach:
the experimenter must select their design while an adversary attempts to select
the least favourable parameterisation. We present theoretical properties of
this approach and show it can be used with gradient based optimisation methods
to find designs efficiently in practice.
| stat |
Elastic Integrative Analysis of Randomized Trial and Real-World Data for
Treatment Heterogeneity Estimation | Parallel randomized trial (RT) and real-world (RW) data are becoming
increasingly available for treatment evaluation. Given the complementary
features of the RT and RW data, we propose a test-based elastic integrative
analysis of the RT and RW data for accurate and robust estimation of the
heterogeneity of treatment effect (HTE), which lies at the heart of precision
medicine. When the RW data are not subject to bias, e.g., due to unmeasured
confounding, our approach combines the RT and RW data for optimal estimation by
exploiting semiparametric efficiency theory. Utilizing the design advantage of
RTs, we construct a built-in test procedure to gauge the reliability of the RW
data and decide whether or not to use RW data in an integrative analysis. We
characterize the asymptotic distribution of the test-based elastic integrative
estimator under local alternatives, which provides a better approximation of
the finite-sample behaviors of the test and estimator when the idealistic
assumption required for the RW data is weakly violated. We provide a
data-adaptive procedure to select the threshold of the test statistic that
promises the smallest mean square error of the proposed estimator of the HTE.
Lastly, we construct an elastic confidence interval that has a good
finite-sample coverage property. We apply the proposed method to characterize
who can benefit from adjuvant chemotherapy in patients with stage IB non-small
cell lung cancer.
| stat |
Equivariant Hamiltonian Flows | This paper introduces equivariant hamiltonian flows, a method for learning
expressive densities that are invariant with respect to a known Lie-algebra of
local symmetry transformations while providing an equivariant representation of
the data. We provide proof of principle demonstrations of how such flows can be
learnt, as well as how the addition of symmetry invariance constraints can
improve data efficiency and generalisation. Finally, we make connections to
disentangled representation learning and show how this work relates to a
recently proposed definition.
| stat |
Sequential Optimization in Locally Important Dimensions | Optimizing an expensive, black-box function $f(\cdot)$ is challenging when
its input space is high-dimensional. Sequential design frameworks first model
$f(\cdot)$ with a surrogate function and then optimize an acquisition function
to determine input settings to evaluate next. Optimization of both $f(\cdot)$
and the acquisition function benefit from effective dimension reduction. Global
variable selection detects and removes input variables that do not affect
$f(\cdot)$ across the input space. Further dimension reduction may be possible
if we consider local variable selection around the current optimum estimate. We
develop a sequential design algorithm called Sequential Optimization in Locally
Important Dimensions (SOLID) that incorporates global and local variable
selection to optimize a continuous, differentiable function. SOLID performs
local variable selection by comparing the surrogate's predictions in a
localized region around the estimated optimum with the $p$ alternative
predictions made by removing each input variable. The search space of the
acquisition function is further restricted to focus only on the variables that
are deemed locally active, leading to greater emphasis on refining the
surrogate model in locally active dimensions. A simulation study across three
test functions and an application to the Sarcos robot dataset show that SOLID
outperforms conventional approaches.
| stat |
Scaled Vecchia approximation for fast computer-model emulation | Many scientific phenomena are studied using computer experiments consisting
of multiple runs of a computer model while varying the input settings. Gaussian
processes (GPs) are a popular tool for the analysis of computer experiments,
enabling interpolation between input settings, but direct GP inference is
computationally infeasible for large datasets. We adapt and extend a powerful
class of GP methods from spatial statistics to enable the scalable analysis and
emulation of large computer experiments. Specifically, we apply Vecchia's
ordered conditional approximation in a transformed input space, with each input
scaled according to how strongly it relates to the computer-model response. The
scaling is learned from the data, by estimating parameters in the GP covariance
function using Fisher scoring. Our methods are highly scalable, enabling
estimation, joint prediction and simulation in near-linear time in the number
of model runs. In several numerical examples, our approach substantially
outperformed existing methods.
| stat |
Split Regularized Regression | We propose an approach for fitting linear regression models that splits the
set of covariates into groups. The optimal split of the variables into groups
and the regularized estimation of the regression coefficients are performed by
minimizing an objective function that encourages sparsity within each group and
diversity among them. The estimated coefficients are then pooled together to
form the final fit. Our procedure works on top of a given penalized linear
regression estimator (e.g., Lasso, elastic net) by fitting it to possibly
overlapping groups of features, encouraging diversity among these groups to
reduce the correlation of the corresponding predictions. For the case of two
groups, elastic net penalty and orthogonal predictors, we give a closed form
solution for the regression coefficients in each group. We establish the
consistency of our method with the number of predictors possibly increasing
with the sample size. An extensive simulation study and real-data applications
show that in general the proposed method improves the prediction accuracy of
the base estimator used in the procedure. Possible extensions to GLMs and other
models are discussed. The supplemental material for this article, available
online, contains the proofs of our theoretical results and the full results of
our simulation study.
| stat |
Learning Rate Adaptation for Federated and Differentially Private
Learning | We propose an algorithm for the adaptation of the learning rate for
stochastic gradient descent (SGD) that avoids the need for validation set use.
The idea for the adaptiveness comes from the technique of extrapolation: to get
an estimate for the error against the gradient flow which underlies SGD, we
compare the result obtained by one full step and two half-steps. The algorithm
is applied in two separate frameworks: federated and differentially private
learning. Using examples of deep neural networks we empirically show that the
adaptive algorithm is competitive with manually tuned commonly used
optimisation methods for differentially privately training. We also show that
it works robustly in the case of federated learning unlike commonly used
optimisation methods.
| stat |
Simultaneous inference for mixed and small area parameters | We address simultaneous inference for mixed parameters which are the key
ingredients in small area estimation. We assume linear mixed model framework.
Firstly, we analyse statistical properties of a max-type statistic and use it
to construct simultaneous prediction intervals as well as to implement multiple
testing procedure. Secondly, we derive bands based on the volume-of-tube
formula. In addition, we adapt some of the simultaneous inference methods from
regression and nonparametric curve estimation and compare them with our
approaches. Simultaneous intervals are necessary to compare clusters since the
presently available intervals are not statistically valid for such analysis.
The proposed testing procedures can be used to validate certain statements
about the set of mixed parameters or to test pairwise differences. Our proposal
is accompanied by simulation experiments and a data example on small area
household incomes. Both of them demonstrate an excellent performance and
utility of our techniques.
| stat |
Nonparametric Density Estimation for High-Dimensional Data - Algorithms
and Applications | Density Estimation is one of the central areas of statistics whose purpose is
to estimate the probability density function underlying the observed data. It
serves as a building block for many tasks in statistical inference,
visualization, and machine learning. Density Estimation is widely adopted in
the domain of unsupervised learning especially for the application of
clustering. As big data become pervasive in almost every area of data sciences,
analyzing high-dimensional data that have many features and variables appears
to be a major focus in both academia and industry. High-dimensional data pose
challenges not only from the theoretical aspects of statistical inference, but
also from the algorithmic/computational considerations of machine learning and
data analytics. This paper reviews a collection of selected nonparametric
density estimation algorithms for high-dimensional data, some of them are
recently published and provide interesting mathematical insights. The important
application domain of nonparametric density estimation, such as { modal
clustering}, are also included in this paper. Several research directions
related to density estimation and high-dimensional data analysis are suggested
by the authors.
| stat |
Graph-based regularization for regression problems with alignment and
highly-correlated designs | Sparse models for high-dimensional linear regression and machine learning
have received substantial attention over the past two decades. Model selection,
or determining which features or covariates are the best explanatory variables,
is critical to the interpretability of a learned model. Much of the current
literature assumes that covariates are only mildly correlated. However, in many
modern applications covariates are highly correlated and do not exhibit key
properties (such as the restricted eigenvalue condition, restricted isometry
property, or other related assumptions). This work considers a high-dimensional
regression setting in which a graph governs both correlations among the
covariates and the similarity among regression coefficients -- meaning there is
\emph{alignment} between the covariates and regression coefficients. Using side
information about the strength of correlations among features, we form a graph
with edge weights corresponding to pairwise covariances. This graph is used to
define a graph total variation regularizer that promotes similar weights for
correlated features.
This work shows how the proposed graph-based regularization yields
mean-squared error guarantees for a broad range of covariance graph structures.
These guarantees are optimal for many specific covariance graphs, including
block and lattice graphs. Our proposed approach outperforms other methods for
highly-correlated design in a variety of experiments on synthetic data and real
biochemistry data.
| stat |
Deep Extreme Value Copulas for Estimation and Sampling | We propose a new method for modeling the distribution function of high
dimensional extreme value distributions. The Pickands dependence function
models the relationship between the covariates in the tails, and we learn this
function using a neural network that is designed to satisfy its required
properties. Moreover, we present new methods for recovering the spectral
representation of extreme distributions and propose a generative model for
sampling from extreme copulas. Numerical examples are provided demonstrating
the efficacy and promise of our proposed methods.
| stat |
Learning with invariances in random features and kernel models | A number of machine learning tasks entail a high degree of invariance: the
data distribution does not change if we act on the data with a certain group of
transformations. For instance, labels of images are invariant under
translations of the images. Certain neural network architectures -- for
instance, convolutional networks -- are believed to owe their success to the
fact that they exploit such invariance properties. With the objective of
quantifying the gain achieved by invariant architectures, we introduce two
classes of models: invariant random features and invariant kernel methods. The
latter includes, as a special case, the neural tangent kernel for convolutional
networks with global average pooling. We consider uniform covariates
distributions on the sphere and hypercube and a general invariant target
function. We characterize the test error of invariant methods in a
high-dimensional regime in which the sample size and number of hidden units
scale as polynomials in the dimension, for a class of groups that we call
`degeneracy $\alpha$', with $\alpha \leq 1$. We show that exploiting invariance
in the architecture saves a $d^\alpha$ factor ($d$ stands for the dimension) in
sample size and number of hidden units to achieve the same test error as for
unstructured architectures.
Finally, we show that output symmetrization of an unstructured kernel
estimator does not give a significant statistical improvement; on the other
hand, data augmentation with an unstructured kernel estimator is equivalent to
an invariant kernel estimator and enjoys the same improvement in statistical
efficiency.
| stat |
Learning Theory for Inferring Interaction Kernels in Second-Order
Interacting Agent Systems | Modeling the complex interactions of systems of particles or agents is a
fundamental scientific and mathematical problem that is studied in diverse
fields, ranging from physics and biology, to economics and machine learning. In
this work, we describe a very general second-order, heterogeneous,
multivariable, interacting agent model, with an environment, that encompasses a
wide variety of known systems. We describe an inference framework that uses
nonparametric regression and approximation theory based techniques to
efficiently derive estimators of the interaction kernels which drive these
dynamical systems. We develop a complete learning theory which establishes
strong consistency and optimal nonparametric min-max rates of convergence for
the estimators, as well as provably accurate predicted trajectories. The
estimators exploit the structure of the equations in order to overcome the
curse of dimensionality and we describe a fundamental coercivity condition on
the inverse problem which ensures that the kernels can be learned and relates
to the minimal singular value of the learning matrix. The numerical algorithm
presented to build the estimators is parallelizable, performs well on
high-dimensional problems, and is demonstrated on complex dynamical systems.
| stat |
Protecting Classifiers From Attacks. A Bayesian Approach | Classification problems in security settings are usually modeled as
confrontations in which an adversary tries to fool a classifier manipulating
the covariates of instances to obtain a benefit. Most approaches to such
problems have focused on game-theoretic ideas with strong underlying common
knowledge assumptions, which are not realistic in the security realm. We
provide an alternative Bayesian framework that accounts for the lack of precise
knowledge about the attacker's behavior using adversarial risk analysis. A key
ingredient required by our framework is the ability to sample from the
distribution of originating instances given the possibly attacked observed one.
We propose a sampling procedure based on approximate Bayesian computation, in
which we simulate the attacker's problem taking into account our uncertainty
about his elements. For large scale problems, we propose an alternative,
scalable approach that could be used when dealing with differentiable
classifiers. Within it, we move the computational load to the training phase,
simulating attacks from an adversary, adapting the framework to obtain a
classifier robustified against attacks.
| stat |
spNNGP R package for Nearest Neighbor Gaussian Process models | This paper describes and illustrates functionality of the spNNGP R package.
The package provides a suite of spatial regression models for Gaussian and
non-Gaussian point-referenced outcomes that are spatially indexed. The package
implements several Markov chain Monte Carlo (MCMC) and MCMC-free Nearest
Neighbor Gaussian Process (NNGP) models for inference about large spatial data.
Non-Gaussian outcomes are modeled using a NNGP Polya-Gamma latent variable.
OpenMP parallelization options are provided to take advantage of multiprocessor
systems. Package features are illustrated using simulated and real data sets.
| stat |
A Wrapped Normal Distribution on Hyperbolic Space for Gradient-Based
Learning | Hyperbolic space is a geometry that is known to be well-suited for
representation learning of data with an underlying hierarchical structure. In
this paper, we present a novel hyperbolic distribution called
\textit{pseudo-hyperbolic Gaussian}, a Gaussian-like distribution on hyperbolic
space whose density can be evaluated analytically and differentiated with
respect to the parameters. Our distribution enables the gradient-based learning
of the probabilistic models on hyperbolic space that could never have been
considered before. Also, we can sample from this hyperbolic probability
distribution without resorting to auxiliary means like rejection sampling. As
applications of our distribution, we develop a hyperbolic-analog of variational
autoencoder and a method of probabilistic word embedding on hyperbolic space.
We demonstrate the efficacy of our distribution on various datasets including
MNIST, Atari 2600 Breakout, and WordNet.
| stat |
Modelling heterogeneity in Latent Space Models for Multidimensional
Networks | Multidimensional network data can have different levels of complexity, as
nodes may be characterized by heterogeneous individual-specific features, which
may vary across the networks. This paper introduces a class of models for
multidimensional network data, where different levels of heterogeneity within
and between networks can be considered. The proposed framework is developed in
the family of latent space models, and it aims to distinguish symmetric
relations between the nodes and node-specific features. Model parameters are
estimated via a Markov Chain Monte Carlo algorithm. Simulated data and an
application to a real example, on fruits import/export data, are used to
illustrate and comment on the performance of the proposed models.
| stat |
A Principle of Least Action for the Training of Neural Networks | Neural networks have been achieving high generalization performance on many
tasks despite being highly over-parameterized. Since classical statistical
learning theory struggles to explain this behavior, much effort has recently
been focused on uncovering the mechanisms behind it, in the hope of developing
a more adequate theoretical framework and having a better control over the
trained models. In this work, we adopt an alternate perspective, viewing the
neural network as a dynamical system displacing input particles over time. We
conduct a series of experiments and, by analyzing the network's behavior
through its displacements, we show the presence of a low kinetic energy
displacement bias in the transport map of the network, and link this bias with
generalization performance. From this observation, we reformulate the learning
problem as follows: finding neural networks which solve the task while
transporting the data as efficiently as possible. This offers a novel
formulation of the learning problem which allows us to provide regularity
results for the solution network, based on Optimal Transport theory. From a
practical viewpoint, this allows us to propose a new learning algorithm, which
automatically adapts to the complexity of the given task, and leads to networks
with a high generalization ability even in low data regimes.
| stat |
$k$-means clustering of extremes | The $k$-means clustering algorithm and its variant, the spherical $k$-means
clustering, are among the most important and popular methods in unsupervised
learning and pattern detection. In this paper, we explore how the spherical
$k$-means algorithm can be applied in the analysis of only the extremal
observations from a data set. By making use of multivariate extreme value
analysis we show how it can be adopted to find "prototypes" of extremal
dependence and we derive a consistency result for our suggested estimator. In
the special case of max-linear models we show furthermore that our procedure
provides an alternative way of statistical inference for this class of models.
Finally, we provide data examples which show that our method is able to find
relevant patterns in extremal observations and allows us to classify extremal
events.
| stat |
Skewed probit regression -- Identifiability, contraction and
reformulation | Skewed probit regression is but one example of a statistical model that
generalizes a simpler model, like probit regression. All skew-symmetric
distributions and link functions arise from symmetric distributions by
incorporating a skewness parameter through some skewing mechanism. In this work
we address some fundamental issues in skewed probit regression, and more
genreally skew-symmetric distributions or skew-symmetric link functions. We
address the issue of identifiability of the skewed probit model parameters by
reformulating the intercept from first principles. A new standardization of the
skew link function is given to provide and anchored interpretation of the
inference. Possible skewness parameters are investigated and the penalizing
complexity priors of these are derived. This prior is invariant under
reparameterization of the skewness parameter and quantifies the contraction of
the skewed probit model to the probit model. The proposed results are available
in the R-INLA package and we illustrate the use and effects of this work using
simulated data, and well-known datasets using the link as well as the
likelihood.
| stat |
Inference in a class of optimization problems: Confidence regions and
finite sample bounds on errors in coverage probabilities | This paper describes a method for carrying out inference on partially
identified parameters that are solutions to a class of optimization problems.
The optimization problems arise in applications in which grouped data are used
for estimation of a model's structural parameters. The parameters are
characterized by restrictions that involve the unknown population means of
observed random variables in addition to the structural parameters of interest.
Inference consists of finding confidence intervals for the structural
parameters. Our theory provides a finite-sample bound on the difference between
the true and nominal probabilities with which a confidence interval contains
the true but unknown value of a parameter. We contrast our method with an
alternative inference method based on the median-of-means estimator of Minsker
(2015). The results of Monte Carlo experiments and empirical examples
illustrate the usefulness of our method.
| stat |
Emulation of stochastic simulators using generalized lambda models | Stochastic simulators are ubiquitous in many fields of applied sciences and
engineering. In the context of uncertainty quantification and optimization, a
large number of simulations are usually necessary, which become intractable for
high-fidelity models. Thus surrogate models of stochastic simulators have been
intensively investigated in the last decade. In this paper, we present a novel
approach to surrogate the response distribution of a stochastic simulator which
uses generalized lambda distributions, whose parameters are represented by
polynomial chaos expansions of the model inputs. As opposed to most existing
approaches, this new method does not require replicated runs of the simulator
at each point of the experimental design. We propose a new fitting procedure
which combines maximum conditional likelihood estimation with (modified)
feasible generalized least-squares. We compare our method with state-of-the-art
nonparametric kernel estimation on four different applications stemming from
mathematical finance and epidemiology. Its performance is illustrated in terms
of the accuracy of both the mean/variance of the stochastic simulator and the
response distribution. As the proposed approach can also be used with
experimental designs containing replications, we carry out a comparison on two
of the examples, that show that replications do not help to get a better
overall accuracy, and may even worsen the results (at fixed total number of
runs of the simulator).
| stat |
Efficiently Breaking the Curse of Horizon in Off-Policy Evaluation with
Double Reinforcement Learning | Off-policy evaluation (OPE) in reinforcement learning is notoriously
difficult in long- and infinite-horizon settings due to diminishing overlap
between behavior and target policies. In this paper, we study the role of
Markovian and time-invariant structure in efficient OPE. We first derive the
efficiency bounds for OPE when one assumes each of these structures. This
precisely characterizes the curse of horizon: in time-variant processes, OPE is
only feasible in the near-on-policy setting, where behavior and target policies
are sufficiently similar. But, in time-invariant Markov decision processes, our
bounds show that truly-off-policy evaluation is feasible, even with only just
one dependent trajectory, and provide the limits of how well we could hope to
do. We develop a new estimator based on Double Reinforcement Learning (DRL)
that leverages this structure for OPE using the efficient influence function we
derive. Our DRL estimator simultaneously uses estimated stationary density
ratios and $q$-functions and remains efficient when both are estimated at slow,
nonparametric rates and remains consistent when either is estimated
consistently. We investigate these properties and the performance benefits of
leveraging the problem structure for more efficient OPE.
| stat |
A Unified Approach to Translate Classical Bandit Algorithms to the
Structured Bandit Setting | We consider a finite-armed structured bandit problem in which mean rewards of
different arms are known functions of a common hidden parameter $\theta^*$.
Since we do not place any restrictions of these functions, the problem setting
subsumes several previously studied frameworks that assume linear or invertible
reward functions. We propose a novel approach to gradually estimate the hidden
$\theta^*$ and use the estimate together with the mean reward functions to
substantially reduce exploration of sub-optimal arms. This approach enables us
to fundamentally generalize any classic bandit algorithm including UCB and
Thompson Sampling to the structured bandit setting. We prove via regret
analysis that our proposed UCB-C algorithm (structured bandit versions of UCB)
pulls only a subset of the sub-optimal arms $O(\log T)$ times while the other
sub-optimal arms (referred to as non-competitive arms) are pulled $O(1)$ times.
As a result, in cases where all sub-optimal arms are non-competitive, which can
happen in many practical scenarios, the proposed algorithms achieve bounded
regret. We also conduct simulations on the Movielens recommendations dataset to
demonstrate the improvement of the proposed algorithms over existing structured
bandit algorithms.
| stat |
A random covariance model for bi-level graphical modeling with
application to resting-state fMRI data | This paper considers a novel problem, bi-level graphical modeling, in which
multiple individual graphical models can be considered as variants of a common
group-level graphical model and inference of both the group- and
individual-level graphical models are of interest. Such problem arises from
many applications including multi-subject neuroimaging and genomics data
analysis. We propose a novel and efficient statistical method, the random
covariance model, to learn the group- and individual-level graphical models
simultaneously. The proposed method can be nicely interpreted as a random
covariance model that mimics the random effects model for mean structures in
linear regression. It accounts for similarity between individual graphical
models, identifies group-level connections that are shared by individuals in
the group, and at the same time infers multiple individual-level networks.
Compared to existing multiple graphical modeling methods that only focus on
individual-level networks, our model learns the group-level structure
underlying the multiple individual networks and enjoys computational efficiency
that is particularly attractive for practical use. We further define a measure
of degrees-of-freedom for the complexity of the model that can be used for
model selection. We demonstrate the asymptotic properties of the method and
show its finite sample performance through simulation studies. Finally, we
apply the proposed method to our motivating clinical data, a multi-subject
resting-state functional magnetic resonance imaging (fMRI) dataset collected
from schizophrenia patients.
| stat |
Efficient Estimation of General Treatment Effects using Neural Networks
with A Diverging Number of Confounders | The estimation of causal effects is a primary goal of behavioral, social,
economic and biomedical sciences. Under the unconfounded treatment assignment
condition, adjustment for confounders requires estimating the nuisance
functions relating outcome and/or treatment to confounders. The conventional
approaches rely on either a parametric or a nonparametric modeling strategy to
approximate the nuisance functions. Parametric methods can introduce serious
bias into casual effect estimation due to possible mis-specification, while
nonparametric estimation suffers from the "curse of dimensionality". This paper
proposes a new unified approach for efficient estimation of treatment effects
using feedforward artificial neural networks when the number of covariates is
allowed to increase with the sample size. We consider a general optimization
framework that includes the average, quantile and asymmetric least squares
treatment effects as special cases. Under this unified setup, we develop a
generalized optimization estimator for the treatment effect with the nuisance
function estimated by neural networks. We further establish the consistency and
asymptotic normality of the proposed estimator and show that it attains the
semiparametric efficiency bound. The proposed methods are illustrated via
simulation studies and a real data application.
| stat |
Regional now- and forecasting for data reported with delay: Towards
surveillance of COVID-19 infections | Governments around the world continue to act to contain and mitigate the
spread of COVID-19. The rapidly evolving situation compels officials and
executives to continuously adapt policies and social distancing measures
depending on the current state of the spread of the disease. In this context,
it is crucial for policymakers to have a firm grasp on what the current state
of the pandemic is as well as to have an idea of how the infective situation is
going to unfold in the next days. However, as in many other situations of
compulsorily-notifiable diseases and beyond, cases are reported with delay to a
central register, with this delay deferring an up-to-date view of the state of
things. We provide a stable tool for monitoring current infection levels as
well as predicting infection numbers in the immediate future at the regional
level. We accomplish this through nowcasting of cases that have not yet been
reported as well as through predictions of future infections. We apply our
model to German data, for which our focus lies in predicting and explain
infectious behavior by district.
| stat |
Domination Number of an Interval Catch Digraph Family and its use for
Testing Uniformity | We consider a special type of interval catch digraph (ICD) family for
one-dimensional data in a randomized setting and propose its use for testing
uniformity. These ICDs are defined with an expansion and a centrality
parameter, hence we will refer to this ICD as parameterized ICD (PICD). We
derive the exact (and asymptotic) distribution of the domination number of this
PICD family when its vertices are from a uniform (and non-uniform) distribution
in one dimension for the entire range of the parameters; thereby determine the
parameters for which the asymptotic distribution is non-degenerate. We observe
jumps (from degeneracy to non-degeneracy or from a non-degenerate distribution
to another) in the asymptotic distribution of the domination number at certain
parameter combinations. We use the domination number for testing uniformity of
data in real line, prove its consistency against certain alternatives, and
compare it with two commonly used tests and three recently proposed tests in
literature and also arc density of this ICD and of another ICD family in terms
of size and power. Based on our extensive Monte Carlo simulations, we
demonstrate that domination number of our PICD has higher power for certain
types of deviations from uniformity compared to other tests.
| stat |
Improved Covariance Matrix Estimator using Shrinkage Transformation and
Random Matrix Theory | One of the major challenges in multivariate analysis is the estimation of
population covariance matrix from sample covariance matrix (SCM). Most recent
covariance matrix estimators use either shrinkage transformations or asymptotic
results from Random Matrix Theory (RMT). Shrinkage techniques help in pulling
extreme correlation values towards certain target values whereas tools from RMT
help in removing noisy eigenvalues of SCM. Both of these techniques use
different approaches to achieve a similar goal which is to remove noisy
correlations and add structure to SCM to overcome the bias-variance trade-off.
In this paper, we first critically evaluate the pros and cons of these two
techniques and then propose an improved estimator which exploits the advantages
of both by taking an optimally weighted convex combination of covariance
matrices estimated by an improved shrinkage transformation and a RMT based
filter. It is a generalized estimator which can adapt to changing sampling
noise conditions in various datasets by performing hyperparameter optimization.
We show the effectiveness of this estimator on the problem of designing a
financial portfolio with minimum risk. We have chosen this problem because the
complex properties of stock market data provide extreme conditions to test the
robustness of a covariance estimator. Using data from four of the world's
largest stock exchanges, we show that our proposed estimator outperforms
existing estimators in minimizing the out-of-sample risk of the portfolio and
hence predicts population statistics more precisely. Since covariance analysis
is a crucial statistical tool, this estimator can be used in a wide range of
machine learning, signal processing and high dimensional pattern recognition
applications.
| stat |
Replacing the do-calculus with Bayes rule | The concept of causality has a controversial history. The question of whether
it is possible to represent and address causal problems with probability
theory, or if fundamentally new mathematics such as the do calculus is required
has been hotly debated, e.g. Pearl (2001) states "the building blocks of our
scientific and everyday knowledge are elementary facts such as "mud does not
cause rain" and "symptoms do not cause disease" and those facts, strangely
enough, cannot be expressed in the vocabulary of probability calculus". This
has lead to a dichotomy between advocates of causal graphical modeling and the
do calculus, and researchers applying Bayesian methods. In this paper we
demonstrate that, while it is critical to explicitly model our assumptions on
the impact of intervening in a system, provided we do so, estimating causal
effects can be done entirely within the standard Bayesian paradigm. The
invariance assumptions underlying causal graphical models can be encoded in
ordinary Probabilistic graphical models, allowing causal estimation with
Bayesian statistics, equivalent to the do calculus. Elucidating the connections
between these approaches is a key step toward enabling the insights provided by
each to be combined to solve real problems.
| stat |
Mixture Hidden Markov Models for Sequence Data: The seqHMM Package in R | Sequence analysis is being more and more widely used for the analysis of
social sequences and other multivariate categorical time series data. However,
it is often complex to describe, visualize, and compare large sequence data,
especially when there are multiple parallel sequences per subject. Hidden
(latent) Markov models (HMMs) are able to detect underlying latent structures
and they can be used in various longitudinal settings: to account for
measurement error, to detect unobservable states, or to compress information
across several types of observations. Extending to mixture hidden Markov models
(MHMMs) allows clustering data into homogeneous subsets, with or without
external covariates.
The seqHMM package in R is designed for the efficient modeling of sequences
and other categorical time series data containing one or multiple subjects with
one or multiple interdependent sequences using HMMs and MHMMs. Also other
restricted variants of the MHMM can be fitted, e.g., latent class models,
Markov models, mixture Markov models, or even ordinary multinomial regression
models with suitable parameterization of the HMM. Good graphical presentations
of data and models are useful during the whole analysis process from the first
glimpse at the data to model fitting and presentation of results. The package
provides easy options for plotting parallel sequence data, and proposes
visualizing HMMs as directed graphs.
| stat |
Kernel Mean Embedding Based Hypothesis Tests for Comparing Spatial Point
Patterns | This paper introduces an approach for detecting differences in the
first-order structures of spatial point patterns. The proposed approach
leverages the kernel mean embedding in a novel way by introducing its
approximate version tailored to spatial point processes. While the original
embedding is infinite-dimensional and implicit, our approximate embedding is
finite-dimensional and comes with explicit closed-form formulas. With its help
we reduce the pattern comparison problem to the comparison of means in the
Euclidean space. Hypothesis testing is based on conducting t-tests on each
dimension of the embedding and combining the resulting p-values using one of
the recently introduced p-value combination techniques. If desired,
corresponding Bayes factors can be computed and averaged over all tests to
quantify the evidence against the null. The main advantages of the proposed
approach are that it can be applied to both single and replicated pattern
comparisons and that neither bootstrap nor permutation procedures are needed to
obtain or calibrate the p-values. Our experiments show that the resulting tests
are powerful and the p-values are well-calibrated; two applications to real
world data are presented.
| stat |
Nonparametric Bayesian volatility estimation | Given discrete time observations over a fixed time interval, we study a
nonparametric Bayesian approach to estimation of the volatility coefficient of
a stochastic differential equation. We postulate a histogram-type prior on the
volatility with piecewise constant realisations on bins forming a partition of
the time interval. The values on the bins are assigned an inverse Gamma Markov
chain (IGMC) prior. Posterior inference is straightforward to implement via
Gibbs sampling, as the full conditional distributions are available explicitly
and turn out to be inverse Gamma. We also discuss in detail the hyperparameter
selection for our method. Our nonparametric Bayesian approach leads to good
practical results in representative simulation examples. Finally, we apply it
on a classical data set in change-point analysis: weekly closings of the
Dow-Jones industrial averages.
| stat |
MIXANDMIX: numerical techniques for the computation of empirical
spectral distributions of population mixtures | The MIXANDMIX (mixtures by Anderson mixing) tool for the computation of the
empirical spectral distribution of random matrices generated by mixtures of
populations is described. Within the population mixture model the mapping
between the population distributions and the limiting spectral distribution can
be obtained by solving a set of systems of non-linear equations, for which an
efficient implementation is provided. The contributions include a method for
accelerated fixed point convergence, a homotopy continuation strategy to
prevent convergence to non-admissible solutions, a blind non-uniform grid
construction for effective distribution support detection and approximation,
and a parallel computing architecture. Comparisons are performed with available
packages for the single population case and with results obtained by simulation
for the more general model implemented here. Results show competitive
performance and improved flexibility.
| stat |
Automatic dimensionality selection for principal component analysis
models with the ignorance score | Principal component analysis (PCA) is by far the most widespread tool for
unsupervised learning with high-dimensional data sets. Its application is
popularly studied for the purpose of exploratory data analysis and online
process monitoring. Unfortunately, fine-tuning PCA models and particularly the
number of components remains a challenging task. Today, this selection is often
based on a combination of guiding principles, experience, and process
understanding. Unlike the case of regression, where cross-validation of the
prediction error is a widespread and trusted approach for model selection,
there are no tools for PCA model selection which reach this level of
acceptance. In this work, we address this challenge and evaluate the utility of
the cross-validated ignorance score with both simulated and experimental data
sets. Application of this method is based on the interpretation of PCA as a
density model, as in probabilistic principal component analysis, and is shown
to be a valuable tool to identify an optimal number of principal components.
| stat |
Finite-sample analysis of interpolating linear classifiers in the
overparameterized regime | We prove bounds on the population risk of the maximum margin algorithm for
two-class linear classification. For linearly separable training data, the
maximum margin algorithm has been shown in previous work to be equivalent to a
limit of training with logistic loss using gradient descent, as the training
error is driven to zero. We analyze this algorithm applied to random data
including misclassification noise. Our assumptions on the clean data include
the case in which the class-conditional distributions are standard normal
distributions. The misclassification noise may be chosen by an adversary,
subject to a limit on the fraction of corrupted labels. Our bounds show that,
with sufficient over-parameterization, the maximum margin algorithm trained on
noisy data can achieve nearly optimal population risk.
| stat |
Misuse of the sign test in narrative synthesis of evidence | In narrative synthesis of evidence, it can be the case that the only
quantitative measures available concerning the efficacy of an intervention is
the direction of the effect, i.e. whether it is positive or negative. In such
situations, the sign test has been proposed in the literature and in recent
Cochrane guidelines as a way to test whether the proportion of positive effects
is favourable. I argue that the sign test is inappropriate in this context as
the data are not generated according to the Binomial distribution it employs. I
demonstrate possible consequences for both hypothesis testing and estimation
via hypothetical examples.
| stat |
Goodness-of-fit tests on manifolds | We develop a general theory for the goodness-of-fit test to non-linear
models. In particular, we assume that the observations are noisy samples of a
submanifold defined by a \yao{sufficiently smooth non-linear map}. The
observation noise is additive Gaussian. Our main result shows that the
"residual" of the model fit, by solving a non-linear least-square problem,
follows a (possibly noncentral) $\chi^2$ distribution. The parameters of the
$\chi^2$ distribution are related to the model order and dimension of the
problem. We further present a method to select the model orders sequentially.
We demonstrate the broad application of the general theory in machine learning
and signal processing, including determining the rank of low-rank (possibly
complex-valued) matrices and tensors from noisy, partial, or indirect
observations, determining the number of sources in signal demixing, and
potential applications in determining the number of hidden nodes in neural
networks.
| stat |
krippendorffsalpha: An R Package for Measuring Agreement Using
Krippendorff's Alpha Coefficient | R package krippendorffsalpha provides tools for measuring agreement using
Krippendorff's Alpha coefficient, a well-known nonparametric measure of
agreement (also called inter-rater reliability and various other names). This
article first develops Krippendorff's Alpha in a natural way, and situates
Alpha among statistical procedures. Then the usage of package
krippendorffsalpha is illustrated via analyses of two datasets, the latter of
which was collected during an imaging study of hip cartilage. The package
permits users to apply the Alpha methodology using built-in distance functions
for the nominal, ordinal, interval, or ratio levels of measurement.
User-defined distance functions are also supported. The fitting function can
accommodate any number of units, any number of coders, and missingness.
Bootstrap inference is supported, and the bootstrap computation can be carried
out in parallel.
| stat |
Recovering Bandits | We study the recovering bandits problem, a variant of the stochastic
multi-armed bandit problem where the expected reward of each arm varies
according to some unknown function of the time since the arm was last played.
While being a natural extension of the classical bandit problem that arises in
many real-world settings, this variation is accompanied by significant
difficulties. In particular, methods need to plan ahead and estimate many more
quantities than in the classical bandit setting. In this work, we explore the
use of Gaussian processes to tackle the estimation and planing problem. We also
discuss different regret definitions that let us quantify the performance of
the methods. To improve computational efficiency of the methods, we provide an
optimistic planning approximation. We complement these discussions with regret
bounds and empirical studies.
| stat |
Multilevel Optimal Transport: a Fast Approximation of Wasserstein-1
distances | We propose a fast algorithm for the calculation of the Wasserstein-1
distance, which is a particular type of optimal transport distance with
homogeneous of degree one transport cost.
Our algorithm is built on multilevel primal-dual algorithms. Several
numerical examples and a complexity analysis are provided to demonstrate its
computational speed. On some commonly used image examples of size
$512\times512$, the proposed algorithm gives solutions within $0.2\sim 1.5$
seconds on a single CPU, which is much faster than the state-of-the-art
algorithms.
| stat |
Dynamic Weights in Gaussian Mixture Models: A Bayesian Approach | In this paper we consider a Gaussian mixture model where the mixture weight
behaves as an unknown function of time. To estimate the mixture weight
function, we develop a Bayesian nonlinear dynamic approach for polynomial
models. Two estimation methods that can be extended to other situations are
considered. One of them, called here component-wise Metropolis-Hastings, is
more general and can be used for any situation where the observation and state
equations are nonlinearly connected. The other method tends to be faster but
must be applied specifically to binary data (by using a probit link function).
This kind of Gaussian mixture model is capable of successfully capturing the
features of the data, as observed in numerical studies. It can be useful in
studies such as clustering, change-point and process control. We apply the
proposed method an array Comparative Genomic Hybridization (aCGH) dataset from
glioblastoma cancer studies, where we illustrate the ability of the new method
to detect chromosome aberrations.
| stat |
Bambi: A simple interface for fitting Bayesian linear models in Python | The popularity of Bayesian statistical methods has increased dramatically in
recent years across many research areas and industrial applications. This is
the result of a variety of methodological advances with faster and cheaper
hardware as well as the development of new software tools. Here we introduce an
open source Python package named Bambi (BAyesian Model Building Interface) that
is built on top of the PyMC3 probabilistic programming framework and the ArviZ
package for exploratory analysis of Bayesian models. Bambi makes it easy to
specify complex generalized linear hierarchical models using a formula notation
similar to those found in the popular R packages lme4, nlme, rstanarm and brms.
We demonstrate Bambi's versatility and ease of use with a few examples spanning
a range of common statistical models including multiple regression, logistic
regression, and mixed-effects modeling with crossed group specific effects.
Additionally we discuss how automatic priors are constructed. Finally, we
conclude with a discussion of our plans for the future development of Bambi.
| stat |
Nonparametric analysis of nonhomogeneous multi-state processes based on
clustered observations | Frequently, clinical trials and observational studies involve complex event
history data with multiple events. When the observations are independent, the
analysis of such studies can be based on standard methods for multi-state
models. However, the independence assumption is often violated, such as in
multicenter studies, which makes the use of standard methods improper. In this
work we address the issue of nonparametric estimation and two-sample testing
for the population-averaged transition and state occupation probabilities under
general multi-state models based on right-censored, left-truncated, and
clustered observations. The proposed methods do not impose assumptions
regarding the within-cluster dependence, allow for informative cluster size,
and are applicable to both Markov and non-Markov processes. Using empirical
process theory, the estimators are shown to be uniformly consistent and to
converge weakly to tight Gaussian processes. Closed-form variance estimators
are derived, rigorous methodology for the calculation of simultaneous
confidence bands is proposed, and the asymptotic properties of the
nonparametric tests are established. Furthermore, we provide theoretical
arguments for the validity of the nonparametric cluster bootstrap, which can be
readily implemented in practice regardless of how complex the underlying
multi-state model is. Simulation studies show that the performance of the
proposed methods is good, and that methods that ignore the within-cluster
dependence can lead to invalid inferences. Finally, the methods are applied to
data from a multicenter randomized controlled trial.
| stat |
An example of application of optimal sample allocation in a finite
population | The problem of estimating a proportion of objects with particular attribute
in a finite population is considered. This paper shows an example of the
application of estimation fraction using new proposed sample allocation in a
population divided into two strata. Variance of estimator of proportion which
uses proposed sample allocation is compared to variance of the standard one.
| stat |
Incremental causal effects | Causal evidence is needed to act and it is often enough for the evidence to
point towards a direction of the effect of an action. For example, policymakers
might be interested in estimating the effect of slightly increasing taxes on
private spending across the whole population. We study identifiability and
estimation of causal effects, where a continuous treatment is slightly shifted
across the whole population (termed average partial effect or incremental
causal effect). We show that incremental effects are identified under local
ignorability and local overlap assumptions, where exchangeability and
positivity only hold in a neighborhood of units. Average treatment effects are
not identified under these assumptions. In this case, and under a smoothness
condition, the incremental effect can be estimated via the average derivative.
Moreover, we prove that in certain finite-sample observational settings,
estimating the incremental effect is easier than estimating the average
treatment effect in terms of asymptotic variance. For high-dimensional
settings, we develop a simple feature transformation that allows for
doubly-robust estimation and inference of incremental causal effects. Finally,
we compare the behaviour of estimators of the incremental treatment effect and
average treatment effect in experiments including data-inspired simulations.
| stat |
Semiparametric counterfactual density estimation | Causal effects are often characterized with averages, which can give an
incomplete picture of the underlying counterfactual distributions. Here we
consider estimating the entire counterfactual density and generic functionals
thereof. We focus on two kinds of target parameters. The first is a density
approximation, defined by a projection onto a finite-dimensional model using a
generalized distance metric, which includes f-divergences as well as $L_p$
norms. The second is the distance between counterfactual densities, which can
be used as a more nuanced effect measure than the mean difference, and as a
tool for model selection. We study nonparametric efficiency bounds for these
targets, giving results for smooth but otherwise generic models and distances.
Importantly, we show how these bounds connect to means of particular
non-trivial functions of counterfactuals, linking the problems of density and
mean estimation. We go on to propose doubly robust-style estimators for the
density approximations and distances, and study their rates of convergence,
showing they can be optimally efficient in large nonparametric models. We also
give analogous methods for model selection and aggregation, when many models
may be available and of interest. Our results all hold for generic models and
distances, but throughout we highlight what happens for particular choices,
such as $L_2$ projections on linear models, and KL projections on exponential
families. Finally we illustrate by estimating the density of CD4 count among
patients with HIV, had all been treated with combination therapy versus
zidovudine alone, as well as a density effect. Our results suggest combination
therapy may have increased CD4 count most for high-risk patients. Our methods
are implemented in the freely available R package npcausal on GitHub.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.