title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Counterfactual Predictions under Runtime Confounding | Algorithms are commonly used to predict outcomes under a particular decision
or intervention, such as predicting whether an offender will succeed on parole
if placed under minimal supervision. Generally, to learn such counterfactual
prediction models from observational data on historical decisions and
corresponding outcomes, one must measure all factors that jointly affect the
outcomes and the decision taken. Motivated by decision support applications, we
study the counterfactual prediction task in the setting where all relevant
factors are captured in the historical data, but it is either undesirable or
impermissible to use some such factors in the prediction model. We refer to
this setting as runtime confounding. We propose a doubly-robust procedure for
learning counterfactual prediction models in this setting. Our theoretical
analysis and experimental results suggest that our method often outperforms
competing approaches. We also present a validation procedure for evaluating the
performance of counterfactual prediction methods.
| stat |
A Bayesian hierarchical model for bridging across patient subgroups in
phase I clinical trials with animal data | Incorporating preclinical animal data, which can be regarded as a special
kind of historical data, into phase I clinical trials can improve decision
making when very little about human toxicity is known. In this paper, we
develop a robust hierarchical modelling approach to leverage animal data into
new phase I clinical trials, where we bridge across non-overlapping,
potentially heterogeneous patient subgroups. Translation parameters are used to
bring both historical and contemporary data onto a common dosing scale. This
leads to feasible exchangeability assumptions that the parameter vectors, which
underpin the dose-toxicity relationship per study, are assumed to be drawn from
a common distribution. Moreover, human dose-toxicity parameter vectors are
assumed to be exchangeable either with the standardised, animal study-specific
parameter vectors, or between themselves. Possibility of non-exchangeability
for each parameter vector is considered to avoid inferences for extreme
subgroups being overly influenced by the other. We illustrate the proposed
approach with several trial data examples, and evaluate the operating
characteristics of our model compared with several alternatives in a simulation
study. Numerical results show that our approach yields robust inferences in
circumstances, where data from multiple sources are inconsistent and/or the
bridging assumptions are incorrect.
| stat |
Federated Learning in Distributed Medical Databases: Meta-Analysis of
Large-Scale Subcortical Brain Data | At this moment, databanks worldwide contain brain images of previously
unimaginable numbers. Combined with developments in data science, these massive
data provide the potential to better understand the genetic underpinnings of
brain diseases. However, different datasets, which are stored at different
institutions, cannot always be shared directly due to privacy and legal
concerns, thus limiting the full exploitation of big data in the study of brain
disorders. Here we propose a federated learning framework for securely
accessing and meta-analyzing any biomedical data without sharing individual
information. We illustrate our framework by investigating brain structural
relationships across diseases and clinical cohorts. The framework is first
tested on synthetic data and then applied to multi-centric, multi-database
studies including ADNI, PPMI, MIRIAD and UK Biobank, showing the potential of
the approach for further applications in distributed analysis of multi-centric
cohorts
| stat |
Artificial Neural Networks to Impute Rounded Zeros in Compositional Data | Methods of deep learning have become increasingly popular in recent years,
but they have not arrived in compositional data analysis. Imputation methods
for compositional data are typically applied on additive, centered or isometric
log-ratio representations of the data. Generally, methods for compositional
data analysis can only be applied to observed positive entries in a data
matrix. Therefore one tries to impute missing values or measurements that were
below a detection limit. In this paper, a new method for imputing rounded zeros
based on artificial neural networks is shown and compared with conventional
methods. We are also interested in the question whether for ANNs, a
representation of the data in log-ratios for imputation purposes, is relevant.
It can be shown, that ANNs are competitive or even performing better when
imputing rounded zeros of data sets with moderate size. They deliver better
results when data sets are big. Also, we can see that log-ratio transformations
within the artificial neural network imputation procedure nevertheless help to
improve the results. This proves that the theory of compositional data analysis
and the fulfillment of all properties of compositional data analysis is still
very important in the age of deep learning.
| stat |
Undirected Graphical Models as Approximate Posteriors | The representation of the approximate posterior is a critical aspect of
effective variational autoencoders (VAEs). Poor choices for the approximate
posterior have a detrimental impact on the generative performance of VAEs due
to the mismatch with the true posterior. We extend the class of posterior
models that may be learned by using undirected graphical models. We develop an
efficient method to train undirected approximate posteriors by showing that the
gradient of the training objective with respect to the parameters of the
undirected posterior can be computed by backpropagation through Markov chain
Monte Carlo updates. We apply these gradient estimators for training discrete
VAEs with Boltzmann machines as approximate posteriors and demonstrate that
undirected models outperform previous results obtained using directed graphical
models. Our implementation is available at https://github.com/QuadrantAI/dvaess .
| stat |
A few statistical principles for data science | In any other circumstance, it might make sense to define the extent of the
terrain (Data Science) first, and then locate and describe the landmarks
(Principles). But this data revolution we are experiencing defies a cadastral
survey. Areas are continually being annexed into Data Science. For example,
biometrics was traditionally statistics for agriculture in all its forms but
now, in Data Science, it means the study of characteristics that can be used to
identify an individual. Examples of non-intrusive measurements include height,
weight, fingerprints, retina scan, voice, photograph/video (facial landmarks
and facial expressions), and gait. A multivariate analysis of such data would
be a complex project for a statistician, but a software engineer might appear
to have no trouble with it at all. In any applied-statistics project, the
statistician worries about uncertainty and quantifies it by modelling data as
realisations generated from a probability space. Another approach to
uncertainty quantification is to find similar data sets, and then use the
variability of results between these data sets to capture the uncertainty. Both
approaches allow 'error bars' to be put on estimates obtained from the original
data set, although the interpretations are different. A third approach, that
concentrates on giving a single answer and gives up on uncertainty
quantification, could be considered as Data Engineering, although it has staked
a claim in the Data Science terrain. This article presents a few (actually
nine) statistical principles for data scientists that have helped me, and
continue to help me, when I work on complex interdisciplinary projects.
| stat |
On Adversarial Mixup Resynthesis | In this paper, we explore new approaches to combining information encoded
within the learned representations of auto-encoders. We explore models that are
capable of combining the attributes of multiple inputs such that a
resynthesised output is trained to fool an adversarial discriminator for real
versus synthesised data. Furthermore, we explore the use of such an
architecture in the context of semi-supervised learning, where we learn a
mixing function whose objective is to produce interpolations of hidden states,
or masked combinations of latent representations that are consistent with a
conditioned class label. We show quantitative and qualitative evidence that
such a formulation is an interesting avenue of research.
| stat |
Bayesian Workflow | The Bayesian approach to data analysis provides a powerful way to handle
uncertainty in all observations, model parameters, and model structure using
probability theory. Probabilistic programming languages make it easier to
specify and fit Bayesian models, but this still leaves us with many options
regarding constructing, evaluating, and using these models, along with many
remaining challenges in computation. Using Bayesian inference to solve
real-world problems requires not only statistical skills, subject matter
knowledge, and programming, but also awareness of the decisions made in the
process of data analysis. All of these aspects can be understood as part of a
tangled workflow of applied Bayesian statistics. Beyond inference, the workflow
also includes iterative model building, model checking, validation and
troubleshooting of computational problems, model understanding, and model
comparison. We review all these aspects of workflow in the context of several
examples, keeping in mind that in practice we will be fitting many models for
any given problem, even if only a subset of them will ultimately be relevant
for our conclusions.
| stat |
Bayesian spatio-temporal model for high-resolution short-term
forecasting of precipitation fields | With extreme weather events becoming more common, the risk posed by surface
water flooding is ever increasing. In this work we propose a model, and
associated Bayesian inference scheme, for generating probabilistic
(high-resolution short-term) forecasts of localised precipitation. The
parametrisation of our underlying hierarchical dynamic spatio-temporal model is
motivated by a forward-time, centred-space finite difference solution to a
collection of stochastic partial differential equations, where the main driving
forces are advection and diffusion. Observations from both weather radar and
ground based rain gauges provide information from which we can learn about the
likely values of the (latent) precipitation field in addition to other unknown
model parameters. Working in the Bayesian paradigm provides a coherent
framework for capturing uncertainty both in the underlying model parameters and
also in our forecasts. Further, appealing to simulation based (MCMC) sampling
yields a straightforward solution to handling zeros, treated as censored
observations, via data augmentation. Both the underlying state and the
observations are of moderately large dimension ($\mathcal{O}(10^4)$ and
$\mathcal{O}(10^3)$ respectively) and this renders standard inference
approaches computationally infeasible. Our solution is to embed the ensemble
Kalman smoother within a Gibbs sampling scheme to facilitate approximate
Bayesian inference in reasonable time. Both the methodology and the
effectiveness of our posterior sampling scheme are demonstrated via simulation
studies and also by a case study of real data from the Urban Observatory
project based in Newcastle upon Tyne, UK.
| stat |
Aggregating Knockoffs for False Discovery Rate Control with an
Application to Gut Microbiome Data | Recent discoveries suggest that our gut microbiome plays an important role in
our health and wellbeing. However, the gut microbiome data are intricate; for
example, the microbial diversity in the gut makes the data high-dimensional.
While there are dedicated high-dimensional methods, such as the lasso
estimator, they always come with the risk of false discoveries. Knockoffs are a
recent approach to control the number of false discoveries. In this paper, we
show that knockoffs can be aggregated to increase power while retaining sharp
control over the false discoveries. We support our method both in theory and
simulations, and we show that it can lead to new discoveries on microbiome data
from the American Gut Project. In particular, our results indicate that several
phyla that have been overlooked so far are associated with obesity.
| stat |
Joint Fairness Model with Applications to Risk Predictions for
Under-represented Populations | Under-representation of certain populations, based on gender, race/ethnicity,
and age, in data collection for predictive modeling may yield less-accurate
predictions for the under-represented groups. Recently, this issue of fairness
in predictions has attracted significant attention, as data-driven models are
increasingly utilized to perform crucial decision-making tasks. Methods to
achieve fairness in the machine learning literature typically build a single
prediction model subject to some fairness criteria in a manner that encourages
fair prediction performances for all groups. These approaches have two major
limitations: i) fairness is often achieved by compromising accuracy for some
groups; ii) the underlying relationship between dependent and independent
variables may not be the same across groups. We propose a Joint Fairness Model
(JFM) approach for binary outcomes that estimates group-specific classifiers
using a joint modeling objective function that incorporates fairness criteria
for prediction. We introduce an Accelerated Smoothing Proximal Gradient
Algorithm to solve the convex objective function, and demonstrate the
properties of the proposed JFM estimates. Next, we presented the key asymptotic
properties for the JFM parameter estimates. We examined the efficacy of the JFM
approach in achieving prediction performances and parities, in comparison with
the Single Fairness Model, group-separate model, and group-ignorant model
through extensive simulations. Finally, we demonstrated the utility of the JFM
method in the motivating example to obtain fair risk predictions for
under-represented older patients diagnosed with coronavirus disease 2019
(COVID-19).
| stat |
A copula-based bivariate integer-valued autoregressive process with
application | A bivariate integer-valued autoregressive process of order 1 (BINAR(1)) with
copula-joint innovations is studied. Different parameter estimation methods are
analyzed and compared via Monte Carlo simulations with emphasis on estimation
of the copula dependence parameter. An empirical application on defaulted and
non-defaulted loan data is carried out using different combinations of copula
functions and marginal distribution functions covering the cases where both
marginal distributions are from the same family, as well as the case where they
are from different distribution families.
| stat |
A general approach to the assessment of uncertainty in volumes by using
the multi-Gaussian model | The goal of this research is to derive an approach to assess uncertainty in
an arbitrary volume conditioned by sampling data, without using geostatistical
simulation. We have accomplished this goal by deriving an numerical tool
suitable for any probabilistic distribution of the sample data. For this, we
have worked with an extension of the traditional multi-Gaussian model, allowing
us to obtain a formulation that makes explicit the dependence of the
uncertainty in the arbitrary volume from the grades within the volume, the
spatial correlation of the data and the conditioning values. A Kriging of the
Gaussian values is the only requirement to obtain not only conditional local
means and variances but also the complete local distributions at any support,
in an easy and straightforward way.
| stat |
Bias detection of $PM_{2.5}$ monitor readings using hidden dynamic
geostatistical calibration model | Accurate and reliable data stream plays an important role in air quality
assessment. Air pollution data collected from monitoring networks, however,
could be biased due to instrumental error or other interventions, which covers
up the real pollution status and misleads policy making. In this study,
motivated by the needs for objective bias detection, we propose a hidden
dynamic geostatistical calibration (HDGC) model to automatically identify
monitoring stations with constantly biased readings. The HDGC model is a
two-level hierarchical model whose parameters are estimated through an
efficient Expectation-Maximization algorithm. Effectiveness of the proposed
model is demonstrated by a simple numerical study. Our method is applied to
hourly $PM_{2.5}$ data from 36 stations in Hebei province, China, over the
period from March 2014 to February 2017. Significantly abnormal readings are
detected from stations in two cities.
| stat |
Know your population and know your model: Using model-based regression
and poststratification to generalize findings beyond the observed sample | Psychology research focuses on interactions, and this has deep implications
for inference from non-representative samples. For the goal of estimating
average treatment effects, we propose to fit a model allowing treatment to
interact with background variables and then average over the distribution of
these variables in the population. This can be seen as an extension of
multilevel regression and poststratification (MRP), a method used in political
science and other areas of survey research, where researchers wish to
generalize from a sparse and possibly non-representative sample to the general
population. In this paper, we discuss areas where this method can be used in
the psychological sciences. We use our method to estimate the norming
distribution for the Big Five Personality Scale using open source data. We
argue that large open data sources like this and other collaborative data
sources can be combined with MRP to help resolve current challenges of
generalizability and replication in psychology.
| stat |
Modeling failures times with dependent renewal type models via
exchangeability | Failure times of a machinery cannot always be assumed independent and
identically distributed, e.g. if after reparations the machinery is not
restored to a same-as-new condition. Framed within the renewal processes
approach, a generalization that considers exchangeable inter-arrival times is
presented. The resulting model provides a more realistic approach to capture
the dependence among events occurring at random times, while retaining much of
the tractability of the classical renewal process. Extensions of some classical
results and special cases of renewal functions are analyzed, in particular the
one corresponding to an exchangeable sequence driven by a Dirichlet process.
The proposal is tested through an estimation procedure using simulated data
sets and with an application to the reliability of hydraulic subsystems in
load-haul-dump machines.
| stat |
Algebraic statistics, tables, and networks: The Fienberg advantage | Stephen Fienberg's affinity for contingency table problems and reinterpreting
models with a fresh look gave rise to a new approach for hypothesis testing of
network models that are linear exponential families. We outline his vision and
influence in this fundamental problem, as well as generalizations to
multigraphs and hypergraphs.
| stat |
Surrogate assisted active subspace and active subspace assisted
surrogate -- A new paradigm for high dimensional structural reliability
analysis | Performing reliability analysis on complex systems is often computationally
expensive. In particular, when dealing with systems having high input
dimensionality, reliability estimation becomes a daunting task. A popular
approach to overcome the problem associated with time-consuming and expensive
evaluations is building a surrogate model. However, these computationally
efficient models often suffer from the curse of dimensionality. Hence, training
a surrogate model for high-dimensional problems is not straightforward.
Henceforth, this paper presents a framework for solving high-dimensional
reliability analysis problems. The basic premise is to train the surrogate
model on a low-dimensional manifold, discovered using the active subspace
algorithm. However, learning the low-dimensional manifold using active subspace
is non-trivial as it requires information on the gradient of the response
variable. To address this issue, we propose using sparse learning algorithms in
conjunction with the active subspace algorithm; the resulting algorithm is
referred to as the sparse active subspace (SAS) algorithm. We project the
high-dimensional inputs onto the identified low-dimensional manifold identified
using SAS. A high-fidelity surrogate model is used to map the inputs on the
low-dimensional manifolds to the output response. We illustrate the efficacy of
the proposed framework by using three benchmark reliability analysis problems
from the literature. The results obtained indicate the accuracy and efficiency
of the proposed approach compared to already established reliability analysis
methods in the literature.
| stat |
Value of Information Analysis via Active Learning and Knowledge Sharing
in Error-Controlled Adaptive Kriging | Large uncertainties in many phenomena have challenged decision making.
Collecting additional information to better characterize reducible
uncertainties is among decision alternatives. Value of information (VoI)
analysis is a mathematical decision framework that quantifies expected
potential benefits of new data and assists with optimal allocation of resources
for information collection. However, analysis of VoI is computational very
costly because of the underlying Bayesian inference especially for
equality-type information. This paper proposes the first surrogate-based
framework for VoI analysis. Instead of modeling the limit state functions
describing events of interest for decision making, which is commonly pursued in
surrogate model-based reliability methods, the proposed framework models system
responses. This approach affords sharing equality-type information from
observations among surrogate models to update likelihoods of multiple events of
interest. Moreover, two knowledge sharing schemes called model and training
points sharing are proposed to most effectively take advantage of the knowledge
offered by costly model evaluations. Both schemes are integrated with an error
rate-based adaptive training approach to efficiently generate accurate Kriging
surrogate models. The proposed VoI analysis framework is applied for an optimal
decision-making problem involving load testing of a truss bridge. While
state-of-the-art methods based on importance sampling and adaptive Kriging
Monte Carlo simulation are unable to solve this problem, the proposed method is
shown to offer accurate and robust estimates of VoI with a limited number of
model evaluations. Therefore, the proposed method facilitates the application
of VoI for complex decision problems.
| stat |
Gradient Descent can Learn Less Over-parameterized Two-layer Neural
Networks on Classification Problems | Recently, several studies have proven the global convergence and
generalization abilities of the gradient descent method for two-layer ReLU
networks. Most studies especially focused on the regression problems with the
squared loss function, except for a few, and the importance of the positivity
of the neural tangent kernel has been pointed out. On the other hand, the
performance of gradient descent on classification problems using the logistic
loss function has not been well studied, and further investigation of this
problem structure is possible. In this work, we demonstrate that the
separability assumption using a neural tangent model is more reasonable than
the positivity condition of the neural tangent kernel and provide a refined
convergence analysis of the gradient descent for two-layer networks with smooth
activations. A remarkable point of our result is that our convergence and
generalization bounds have much better dependence on the network width in
comparison to related studies. Consequently, our theory provides a
generalization guarantee for less over-parameterized two-layer networks, while
most studies require much higher over-parameterization.
| stat |
Modeling space-time trends and dependence in extreme precipitations of
Burkina Faso by the approach of the Peaks-Over-Threshold | Modeling extremes of climate variables in the framework of climate change is
a particularly difficult task, since it implies taking into account
spatio-temporal nonstationarities. In this paper, we propose a new method for
estimating extreme precipitation at the points where we have not observations
using information from marginal distributions and dependence structure. To
reach this goal we combine two statistical approaches of extreme values theory
allowing on the one hand to control temporal and spatial non-stationarities via
a tail trend function with a spatio-temporal structure in the marginal
distributions and by modeling on the other hand the dependence structure by a
latent spatial process using generalized `-Pareto processes. This new
methodology for trend analysis of extreme events is applied to rainfall data
from Burkina Faso. We show that extreme precipitation is spatially and
temporally correlated for distances of approximately 200 km. Locally, extreme
rainfall has more of an upward than downward trend.
| stat |
Doubly-Robust Lasso Bandit | Contextual multi-armed bandit algorithms are widely used in sequential
decision tasks such as news article recommendation systems, web page ad
placement algorithms, and mobile health. Most of the existing algorithms have
regret proportional to a polynomial function of the context dimension, $d$. In
many applications however, it is often the case that contexts are
high-dimensional with only a sparse subset of size $s_0 (\ll d)$ being
correlated with the reward. We consider the stochastic linear contextual bandit
problem and propose a novel algorithm, namely the Doubly-Robust Lasso Bandit
algorithm, which exploits the sparse structure of the regression parameter as
in Lasso, while blending the doubly-robust technique used in missing data
literature. The high-probability upper bound of the regret incurred by the
proposed algorithm does not depend on the number of arms and scales with
$\mathrm{log}(d)$ instead of a polynomial function of $d$. The proposed
algorithm shows good performance when contexts of different arms are correlated
and requires less tuning parameters than existing methods.
| stat |
A meta-inference framework to integrate multiple external models into a
current study | It is becoming increasingly common for researchers to consider incorporating
external information from large studies to improve the accuracy of statistical
inference instead of relying on a modestly sized dataset collected internally.
With some new predictors only available internally, we aim to build improved
regression models based on individual-level data from an "internal" study while
incorporating summary-level information from "external" models. We propose a
meta-analysis framework along with two weighted estimators as the composite of
empirical Bayes estimators, which combines the estimates from the different
external models. The proposed framework is flexible and robust in the ways that
(i) it is capable of incorporating external models that use a slightly
different set of covariates; (ii) it can identify the most relevant external
information and diminish the influence of information that is less compatible
with the internal data; and (iii) it nicely balances the bias-variance
trade-off while preserving the most efficiency gain. The proposed estimators
are more efficient than the naive analysis of the internal data and other naive
combinations of external estimators.
| stat |
An AI model for Rapid and Accurate Identification of Chemical Agents in
Mass Casualty Incidents | In this report we examine the effectiveness of WISER in identification of a
chemical culprit during a chemical based Mass Casualty Incident (MCI). We also
evaluate and compare Binary Decision Tree (BDT) and Artificial Neural Networks
(ANN) using the same experimental conditions as WISER. The reverse engineered
set of Signs/Symptoms from the WISER application was used as the training set
and 31,100 simulated patient records were used as the testing set. Three sets
of simulated patient records were generated by 5%, 10% and 15% perturbation of
the Signs/Symptoms of each chemical record. While all three methods achieved a
100% training accuracy, WISER, BDT and ANN produced performances in the range
of: 1.8%-0%, 65%-26%, 67%-21% respectively. A preliminary investigation of
dimensional reduction using ANN illustrated a dimensional collapse from 79
variables to 40 with little loss of classification performance.
| stat |
Finite Sample Hypothesis Tests for Stacked Estimating Equations | Suppose there are two unknown parameters, each parameter is the solution to
an estimating equation, and the estimating equation of one parameter depends on
the other parameter. The parameters can be jointly estimated by "stacking"
their estimating equations and solving for both parameters simultaneously.
Asymptotic confidence intervals are readily available for stacked estimating
equations. We introduce a bootstrap-based hypothesis test for stacked
estimating equations which does not rely on asymptotic approximations. Test
statistics are constructed by splitting the sample in two, estimating the first
parameter on a portion of the sample then plugging the result into the second
estimating equation to solve for the next parameter using the remaining sample.
To reduce simulation variability from a single split, we repeatedly split the
sample and take the sample mean of all the estimates. For parametric models, we
derive the limiting distribution of sample splitting estimator and show they
are equivalent to stacked estimating equations.
| stat |
Explained Variation under the Additive Hazards Model | We study explained variation under the additive hazards regression model for
right-censored data. We consider different approaches for developing such a
measure, and focus on one that estimates the proportion of variation in the
failure time explained by the covariates. We study the properties of the
measure both analytically, and through extensive simulations. We apply the
measure to a well-known survival data set as well as the linked Surveillance,
Epidemiology and End Results (SEER)-Medicare database for prediction of
mortality in early-stage prostate cancer patients using high dimensional claims
codes.
| stat |
A new paradox in random-effects meta-analysis | Meta-analysis is an important tool for combining results from multiple
studies and has been widely used in evidence-based medicine for several
decades. This paper reports, for the first time, an interesting and valuable
paradox in random-effects meta-analysis that is likely to occur when the number
of studies is small and/or the heterogeneity is large. With the incredible
paradox, we hence advocate meta-analysts to be extremely cautious when
interpreting the final results from the random-effects meta-analysis. And more
importantly, with the unexpected dilemma in making decisions, the new paradox
has raised an open question whether the current random-effects model is
reasonable and tenable for meta-analysis, or it needs to be abandoned or
further improved to some extent.
| stat |
Explaining Visual Models by Causal Attribution | Model explanations based on pure observational data cannot compute the
effects of features reliably, due to their inability to estimate how each
factor alteration could affect the rest. We argue that explanations should be
based on the causal model of the data and the derived intervened causal models,
that represent the data distribution subject to interventions. With these
models, we can compute counterfactuals, new samples that will inform us how the
model reacts to feature changes on our input. We propose a novel explanation
methodology based on Causal Counterfactuals and identify the limitations of
current Image Generative Models in their application to counterfactual
creation.
| stat |
Modelling excess zeros in count data: A new perspective on modelling
approaches | We consider models underlying regression analysis of count data in which the
observed frequency of zero counts is unusually large, typically with respect to
the Poisson distribution. We focus on two alternative modelling approaches:
Over-Dispersion (OD) models, and Zero-Inflation (ZI) models, both of which can
be seen as generalisations of the Poisson distribution; we refer to these as
Implicit and Explicit ZI models, respectively. Although sometimes seen as
competing approaches, they can be complementary; OD is a consequence of ZI
modelling, and ZI is a by-product of OD modelling. The central objective in
such analyses is often concerned with inference on the effect of covariates on
the mean, in light of the excess of zeros in the counts. The contribution of
our paper is to focus on models for different types of ZI, some of which can
only be generated by explicit ZI modelling; and on their characterisation by
considering the induced probability of a zero as a function of the zero
probability of a base distribution (usually Poisson). We develop the underlying
theory for univariate counts. The perspective highlights some of the
difficulties encountered in distinguishing the alternative modelling options.
| stat |
The Inverse G-Wishart Distribution and Variational Message Passing | Message passing on a factor graph is a powerful paradigm for the coding of
approximate inference algorithms for arbitrarily graphical large models. The
notion of a factor graph fragment allows for compartmentalization of algebra
and computer code. We show that the Inverse G-Wishart family of distributions
enables fundamental variational message passing factor graph fragments to be
expressed elegantly and succinctly. Such fragments arise in models for which
approximate inference concerning covariance matrix or variance parameters is
made, and are ubiquitous in contemporary statistics and machine learning.
| stat |
Flexible signal denoising via flexible empirical Bayes shrinkage | Signal denoising---also known as non-parametric regression---is often
performed through shrinkage estimation in a transformed (e.g., wavelet) domain;
shrinkage in the transformed domain corresponds to smoothing in the original
domain. A key question in such applications is how much to shrink, or,
equivalently, how much to smooth. Empirical Bayes shrinkage methods provide an
attractive solution to this problem; they use the data to estimate a
distribution of underlying "effects", hence automatically select an appropriate
amount of shrinkage. However, most existing implementations of Empirical Bayes
shrinkage are less flexible than they could be--both in their assumptions on
the underlying distribution of effects, and in their ability to handle
heterskedasticity---which limits their signal denoising applications. Here we
address this by taking a particularly flexible, stable and computationally
convenient Empirical Bayes shrinkage method, and we apply it to several signal
denoising problems. These applications include smoothing of Poisson data and
heteroskedastic Gaussian data. We show through empirical comparisons that the
results are competitive with other methods, including both simple thresholding
rules and purpose-built Empirical Bayes procedures. Our methods are implemented
in the R package smashr, "SMoothing by Adaptive SHrinkage in R," available at
https://www.github.com/stephenslab/smashr
| stat |
DeepKriging: Spatially Dependent Deep Neural Networks for Spatial
Prediction | In spatial statistics, a common objective is to predict the values of a
spatial process at unobserved locations by exploiting spatial dependence. In
geostatistics, Kriging provides the best linear unbiased predictor using
covariance functions and is often associated with Gaussian processes. However,
when considering non-linear prediction for non-Gaussian and categorical data,
the Kriging prediction is not necessarily optimal, and the associated variance
is often overly optimistic. We propose to use deep neural networks (DNNs) for
spatial prediction. Although DNNs are widely used for general classification
and prediction, they have not been studied thoroughly for data with spatial
dependence. In this work, we propose a novel neural network structure for
spatial prediction by adding an embedding layer of spatial coordinates with
basis functions. We show in theory that the proposed DeepKriging method has
multiple advantages over Kriging and classical DNNs only with spatial
coordinates as features. We also provide density prediction for uncertainty
quantification without any distributional assumption and apply the method to
PM$_{2.5}$ concentrations across the continental United States.
| stat |
A Generalized Neyman-Pearson Criterion for Optimal Domain Adaptation | In the problem of domain adaptation for binary classification, the learner is
presented with labeled examples from a source domain, and must correctly
classify unlabeled examples from a target domain, which may differ from the
source. Previous work on this problem has assumed that the performance measure
of interest is the expected value of some loss function. We introduce a new
Neyman-Pearson-like criterion and argue that, for this optimality criterion,
stronger domain adaptation results are possible than what has previously been
established. In particular, we study a class of domain adaptation problems that
generalizes both the covariate shift assumption and a model for
feature-dependent label noise, and establish optimal classification on the
target domain despite not having access to labelled data from this domain.
| stat |
Heterogeneous Treatment and Spillover Effects under Clustered Network
Interference | The bulk of causal inference studies rules out the presence of interference
between units. However, in many real-world settings units are interconnected by
social, physical or virtual ties and the effect of a treatment can spill from
one unit to other connected individuals in the network. In these settings,
interference should be taken into account to avoid biased estimates of the
treatment effect, but it can also be leveraged to save resources and provide
the intervention to a lower percentage of the population where the treatment is
more effective and where the effect can spill over to other susceptible
individuals. In fact, different people might respond differently not only to
the treatment received but also to the treatment received by their network
contacts. Understanding the heterogeneity of treatment and spillover effects
can help policy-makers in the scale-up phase of the intervention, it can guide
the design of targeting strategies with the ultimate goal of making the
interventions more cost-effective, and it might even allow generalizing the
level of treatment spillover effects in other populations. In this paper, we
develop a machine learning method that makes use of tree-based algorithms and
an Horvitz-Thompson estimator to assess the heterogeneity of treatment and
spillover effects with respect to individual, neighborhood and network
characteristics in the context of clustered network interference. We illustrate
how the proposed binary tree methodology performs in a Monte Carlo simulation
study. Additionally, we provide an application on a randomized experiment aimed
at assessing the heterogeneous effects of information sessions on the uptake of
a new weather insurance policy in rural China.
| stat |
Adjusted Measures for Feature Selection Stability for Data Sets with
Similar Features | For data sets with similar features, for example highly correlated features,
most existing stability measures behave in an undesired way: They consider
features that are almost identical but have different identifiers as different
features. Existing adjusted stability measures, that is, stability measures
that take into account the similarities between features, have major
theoretical drawbacks. We introduce new adjusted stability measures that
overcome these drawbacks. We compare them to each other and to existing
stability measures based on both artificial and real sets of selected features.
Based on the results, we suggest using one new stability measure that considers
highly similar features as exchangeable.
| stat |
Bayesian bandwidth estimation and semi-metric selection for a functional
partial linear model with unknown error density | This study examines the optimal selections of bandwidth and semi-metric for a
functional partial linear model. Our proposed method begins by estimating the
unknown error density using a kernel density estimator of residuals, where the
regression function, consisting of parametric and nonparametric components, can
be estimated by functional principal component and functional Nadayara-Watson
estimators. The estimation accuracy of the regression function and error
density crucially depends on the optimal estimations of bandwidth and
semi-metric. A Bayesian method is utilized to simultaneously estimate the
bandwidths in the regression function and kernel error density by minimizing
the Kullback-Leibler divergence. For estimating the regression function and
error density, a series of simulation studies demonstrate that the functional
partial linear model gives improved estimation and forecast accuracies compared
with the functional principal component regression and functional nonparametric
regression. Using a spectroscopy dataset, the functional partial linear model
yields better forecast accuracy than some commonly used functional regression
models. As a by-product of the Bayesian method, a pointwise prediction interval
can be obtained, and marginal likelihood can be used to select the optimal
semi-metric.
| stat |
Provably robust estimation of modulo 1 samples of a smooth function with
applications to phase unwrapping | Consider an unknown smooth function $f: [0,1]^d \rightarrow \mathbb{R}$, and
say we are given $n$ noisy mod 1 samples of $f$, i.e., $y_i = (f(x_i) +
\eta_i)\mod 1$, for $x_i \in [0,1]^d$, where $\eta_i$ denotes the noise. Given
the samples $(x_i,y_i)_{i=1}^{n}$, our goal is to recover smooth, robust
estimates of the clean samples $f(x_i) \bmod 1$. We formulate a natural
approach for solving this problem, which works with angular embeddings of the
noisy mod 1 samples over the unit circle, inspired by the angular
synchronization framework. This amounts to solving a smoothness regularized
least-squares problem -- a quadratically constrained quadratic program (QCQP)
-- where the variables are constrained to lie on the unit circle. Our approach
is based on solving its relaxation, which is a trust-region sub-problem and
hence solvable efficiently. We provide theoretical guarantees demonstrating its
robustness to noise for adversarial, and random Gaussian and Bernoulli noise
models. To the best of our knowledge, these are the first such theoretical
results for this problem. We demonstrate the robustness and efficiency of our
approach via extensive numerical simulations on synthetic data, along with a
simple least-squares solution for the unwrapping stage, that recovers the
original samples of $f$ (up to a global shift). It is shown to perform well at
high levels of noise, when taking as input the denoised modulo $1$ samples.
Finally, we also consider two other approaches for denoising the modulo 1
samples that leverage tools from Riemannian optimization on manifolds,
including a Burer-Monteiro approach for a semidefinite programming relaxation
of our formulation. For the two-dimensional version of the problem, which has
applications in radar interferometry, we are able to solve instances of
real-world data with a million sample points in under 10 seconds, on a personal
laptop.
| stat |
Imagining The Road Ahead: Multi-Agent Trajectory Prediction via
Differentiable Simulation | We develop a deep generative model built on a fully differentiable simulator
for multi-agent trajectory prediction. Agents are modeled with conditional
recurrent variational neural networks (CVRNNs), which take as input an
ego-centric birdview image representing the current state of the world and
output an action, consisting of steering and acceleration, which is used to
derive the subsequent agent state using a kinematic bicycle model. The full
simulation state is then differentiably rendered for each agent, initiating the
next time step. We achieve state-of-the-art results on the INTERACTION dataset,
using standard neural architectures and a standard variational training
objective, producing realistic multi-modal predictions without any ad-hoc
diversity-inducing losses. We conduct ablation studies to examine individual
components of the simulator, finding that both the kinematic bicycle model and
the continuous feedback from the birdview image are crucial for achieving this
level of performance. We name our model ITRA, for "Imagining the Road Ahead".
| stat |
Duality in RKHSs with Infinite Dimensional Outputs: Application to
Robust Losses | Operator-Valued Kernels (OVKs) and associated vector-valued Reproducing
Kernel Hilbert Spaces provide an elegant way to extend scalar kernel methods
when the output space is a Hilbert space. Although primarily used in finite
dimension for problems like multi-task regression, the ability of this
framework to deal with infinite dimensional output spaces unlocks many more
applications, such as functional regression, structured output prediction, and
structured data representation. However, these sophisticated schemes crucially
rely on the kernel trick in the output space, so that most of previous works
have focused on the square norm loss function, completely neglecting robustness
issues that may arise in such surrogate problems. To overcome this limitation,
this paper develops a duality approach that allows to solve OVK machines for a
wide range of loss functions. The infinite dimensional Lagrange multipliers are
handled through a Double Representer Theorem, and algorithms for
$\epsilon$-insensitive losses and the Huber loss are thoroughly detailed.
Robustness benefits are emphasized by a theoretical stability analysis, as well
as empirical improvements on structured data applications.
| stat |
Sparse Partial Least Squares for Coarse Noisy Graph Alignment | Graph signal processing (GSP) provides a powerful framework for analyzing
signals arising in a variety of domains. In many applications of GSP, multiple
network structures are available, each of which captures different aspects of
the same underlying phenomenon. To integrate these different data sources,
graph alignment techniques attempt to find the best correspondence between
vertices of two graphs. We consider a generalization of this problem, where
there is no natural one-to-one mapping between vertices, but where there is
correspondence between the community structures of each graph. Because we seek
to learn structure at this higher community level, we refer to this problem as
"coarse" graph alignment. To this end, we propose a novel regularized partial
least squares method which both incorporates the observed graph structures and
imposes sparsity in order to reflect the underlying block community structure.
We provide efficient algorithms for our method and demonstrate its
effectiveness in simulations.
| stat |
A Class of Multilevel Nonregular Designs for Studying Quantitative
Factors | Fractional factorial designs are widely used for designing screening
experiments. Nonregular fractional factorial designs can have better properties
than regular designs, but their construction is challenging. Current research
on the construction of nonregular designs focuses on two-level designs. We
provide a novel class of multilevel nonregular designs by permuting levels of
regular designs. We develop a theory illustrating how levels can be permuted
without computer search and accordingly propose a sequential method for
constructing nonregular designs. Compared to regular designs, these nonregular
designs can provide more accurate estimations on factorial effects and more
efficient screening for experiments with quantitative factors. We further
explore the space-filling property of the obtained designs and demonstrate
their superiority.
| stat |
New approach for stochastic downscaling and bias correction of daily
mean temperatures to a high-resolution grid | In applications of climate information, coarse-resolution climate projections
commonly need to be downscaled to a finer grid. One challenge of this
requirement is the modeling of sub-grid variability and the spatial and
temporal dependence at the finer scale. Here, a post-processing procedure is
proposed for temperature projections that addresses this challenge. The
procedure employs statistical bias correction and stochastic downscaling in two
steps. In a first step, errors that are related to spatial and temporal
features of the first two moments of the temperature distribution at model
scale are identified and corrected. Secondly, residual space-time dependence at
the finer scale is analyzed using a statistical model, from which realizations
are generated and then combined with appropriate climate change signal to form
the downscaled projection fields. Using a high-resolution observational gridded
data product, the proposed approach is applied in a case study where
projections of two regional climate models from the EURO-CORDEX ensemble are
bias-corrected and downscaled to a 1x1 km grid in the Trondelag area of Norway.
A cross-validation study shows that the proposed procedure generates results
that better reflect the marginal distributional properties of the data product
and have better consistency in space and time than empirical quantile mapping.
| stat |
Metrics for Multi-Class Classification: an Overview | Classification tasks in machine learning involving more than two classes are
known by the name of "multi-class classification". Performance indicators are
very useful when the aim is to evaluate and compare different classification
models or machine learning techniques. Many metrics come in handy to test the
ability of a multi-class classifier. Those metrics turn out to be useful at
different stage of the development process, e.g. comparing the performance of
two different models or analysing the behaviour of the same model by tuning
different parameters. In this white paper we review a list of the most
promising multi-class metrics, we highlight their advantages and disadvantages
and show their possible usages during the development of a classification
model.
| stat |
A Note on Exploratory Item Factor Analysis by Singular Value
Decomposition | We revisit a singular value decomposition (SVD) algorithm given in Chen et
al. (2019b) for exploratory Item Factor Analysis (IFA). This algorithm
estimates a multidimensional IFA model by SVD and was used to obtain a starting
point for joint maximum likelihood estimation in Chen et al. (2019b). Thanks to
the analytic and computational properties of SVD, this algorithm guarantees a
unique solution and has computational advantage over other exploratory IFA
methods. Its computational advantage becomes significant when the numbers of
respondents, items, and factors are all large. This algorithm can be viewed as
a generalization of principal component analysis (PCA) to binary data. In this
note, we provide the statistical underpinning of the algorithm. In particular,
we show its statistical consistency under the same double asymptotic setting as
in Chen et al. (2019b). We also demonstrate how this algorithm provides a scree
plot for investigating the number of factors and provide its asymptotic theory.
Further extensions of the algorithm are discussed. Finally, simulation studies
suggest that the algorithm has good finite sample performance.
| stat |
Data-driven Optimal Cost Selection for Distributionally Robust
Optimization | Recently, (Blanchet, Kang, and Murhy 2016, and Blanchet, and Kang 2017)
showed that several machine learning algorithms, such as square-root Lasso,
Support Vector Machines, and regularized logistic regression, among many
others, can be represented exactly as distributionally robust optimization
(DRO) problems. The distributional uncertainty is defined as a neighborhood
centered at the empirical distribution. We propose a methodology which learns
such neighborhood in a natural data-driven way. We show rigorously that our
framework encompasses adaptive regularization as a particular case. Moreover,
we demonstrate empirically that our proposed methodology is able to improve
upon a wide range of popular machine learning estimators.
| stat |
Variational Characterizations of Local Entropy and Heat Regularization
in Deep Learning | The aim of this paper is to provide new theoretical and computational
understanding on two loss regularizations employed in deep learning, known as
local entropy and heat regularization. For both regularized losses we introduce
variational characterizations that naturally suggest a two-step scheme for
their optimization, based on the iterative shift of a probability density and
the calculation of a best Gaussian approximation in Kullback-Leibler
divergence. Under this unified light, the optimization schemes for local
entropy and heat regularized loss differ only over which argument of the
Kullback-Leibler divergence is used to find the best Gaussian approximation.
Local entropy corresponds to minimizing over the second argument, and the
solution is given by moment matching. This allows to replace traditional
back-propagation calculation of gradients by sampling algorithms, opening an
avenue for gradient-free, parallelizable training of neural networks.
| stat |
Strata-based Quantification of Distributional Uncertainty in
Socio-Economic Indicators: A Comparative Study of Indian States | This paper reports a comprehensive study of distributional uncertainty in a
few socio-economic indicators across the various states of India over the years
2001-2011. We show that the DGB distribution, a typical rank order
distribution, provide excellent fits to the district-wise empirical data for
the population size, literacy rate (LR) and work participation rate (WPR)
within every states in India, through its two distributional parameters.
Moreover, taking resort to the entropy formulation of the DGB distribution, a
proposed uncertainty percentage (UP) unveils the dynamics of the uncertainty of
LR and WPR in all states of India. We have also commented on the changes in the
estimated parameters and the UP values from the years 2001 to 2011.
Additionally, a gender based analysis of the distribution of these important
socio-economic variables within different states of India has also been
discussed. Interestingly, it has been observed that, although the distributions
of the numbers of literate and working people has a direct (linear)
correspondence with that of the population size, the literacy and
work-participation rates are distributed independently of the population
distributions.
| stat |
Quantized Variational Inference | We present Quantized Variational Inference, a new algorithm for Evidence
Lower Bound maximization. We show how Optimal Voronoi Tesselation produces
variance free gradients for ELBO optimization at the cost of introducing
asymptotically decaying bias. Subsequently, we propose a Richardson
extrapolation type method to improve the asymptotic bound. We show that using
the Quantized Variational Inference framework leads to fast convergence for
both score function and the reparametrized gradient estimator at a comparable
computational cost. Finally, we propose several experiments to assess the
performance of our method and its limitations.
| stat |
Spatial Statistical Models: an overview under the Bayesian Approach | Spatial documentation is exponentially increasing given the availability of
Big IoT Data, enabled by the devices miniaturization and data storage capacity.
Bayesian spatial statistics is a useful statistical tool to determine the
dependence structure and hidden patterns over space through prior knowledge and
data likelihood. Nevertheless, this modeling class is not well explored as the
classification and regression machine learning models given their simplicity
and often weak (data) independence supposition. In this manner, this systematic
review aimed to unravel the main models presented in the literature in the past
20 years, identify gaps, and research opportunities. Elements such as random
fields, spatial domains, prior specification, covariance function, and
numerical approximations were discussed. This work explored the two subclasses
of spatial smoothing global and local.
| stat |
Inference of Dynamic Graph Changes for Functional Connectome | Dynamic functional connectivity is an effective measure for the brain's
responses to continuous stimuli. We propose an inferential method to detect the
dynamic changes of brain networks based on time-varying graphical models.
Whereas most existing methods focus on testing the existence of change points,
the dynamics in the brain network offer more signals in many neuroscience
studies. We propose a novel method to conduct hypothesis testing on changes in
dynamic brain networks. We introduce a bootstrap statistic to approximate the
supreme of the high-dimensional empirical processes over dynamically changing
edges. Our simulations show that this framework can capture the change points
with changed connectivity. Finally, we apply our method to a brain imaging
dataset under a natural audio-video stimulus and illustrate that we are able to
detect temporal changes in brain networks. The functions of the identified
regions are consistent with specific emotional annotations, which are closely
associated with changes inferred by our method.
| stat |
Combining data assimilation and machine learning to emulate a dynamical
model from sparse and noisy observations: a case study with the Lorenz 96
model | A novel method, based on the combination of data assimilation and machine
learning is introduced. The new hybrid approach is designed for a two-fold
scope: (i) emulating hidden, possibly chaotic, dynamics and (ii) predicting
their future states. The method consists in applying iteratively a data
assimilation step, here an ensemble Kalman filter, and a neural network. Data
assimilation is used to optimally combine a surrogate model with sparse noisy
data. The output analysis is spatially complete and is used as a training set
by the neural network to update the surrogate model. The two steps are then
repeated iteratively. Numerical experiments have been carried out using the
chaotic 40-variables Lorenz 96 model, proving both convergence and statistical
skill of the proposed hybrid approach. The surrogate model shows short-term
forecast skill up to two Lyapunov times, the retrieval of positive Lyapunov
exponents as well as the more energetic frequencies of the power density
spectrum. The sensitivity of the method to critical setup parameters is also
presented: the forecast skill decreases smoothly with increased observational
noise but drops abruptly if less than half of the model domain is observed. The
successful synergy between data assimilation and machine learning, proven here
with a low-dimensional system, encourages further investigation of such hybrids
with more sophisticated dynamics.
| stat |
The Hierarchy of Block Models | There exist various types of network block models such as the Stochastic
Block Model (SBM), the Degree Corrected Block Model (DCBM), and the Popularity
Adjusted Block Model (PABM). While this leads to a variety of choices, the
block models do not have a nested structure. In addition, there is a
substantial jump in the number of parameters from the DCBM to the PABM. The
objective of this paper is formulation of a hierarchy of block model which does
not rely on arbitrary identifiability conditions. We propose a Nested Block
Model (NBM) that treats the SBM, the DCBM and the PABM as its particular cases
with specific parameter values, and, in addition, allows a multitude of
versions that are more complicated than DCBM but have fewer unknown parameters
than the PABM. The latter allows one to carry out clustering and estimation
without preliminary testing, to see which block model is really true.
| stat |
Bootstrap independence test for functional linear models | Functional data have been the subject of many research works over the last
years. Functional regression is one of the most discussed issues. Specifically,
significant advances have been made for functional linear regression models
with scalar response. Let $(\mathcal{H},<\cdot,\cdot>)$ be a separable Hilbert
space. We focus on the model $Y=<\Theta,X>+b+\varepsilon$, where $Y$ and
$\varepsilon$ are real random variables, $X$ is an $\mathcal{H}$-valued random
element, and the model parameters $b$ and $\Theta$ are in $\mathbb{R}$ and
$\mathcal{H}$, respectively. Furthermore, the error satisfies that
$E(\varepsilon|X)=0$ and $E(\varepsilon^2|X)=\sigma^2<\infty$. A consistent
bootstrap method to calibrate the distribution of statistics for testing $H_0:
\Theta=0$ versus $H_1: \Theta\neq 0$ is developed. The asymptotic theory, as
well as a simulation study and a real data application illustrating the
usefulness of our proposed bootstrap in practice, is presented.
| stat |
Optimal experimental design via Bayesian optimization: active causal
structure learning for Gaussian process networks | We study the problem of causal discovery through targeted interventions.
Starting from few observational measurements, we follow a Bayesian active
learning approach to perform those experiments which, in expectation with
respect to the current model, are maximally informative about the underlying
causal structure. Unlike previous work, we consider the setting of continuous
random variables with non-linear functional relationships, modelled with
Gaussian process priors. To address the arising problem of choosing from an
uncountable set of possible interventions, we propose to use Bayesian
optimisation to efficiently maximise a Monte Carlo estimate of the expected
information gain.
| stat |
Convergence Analysis of Riemannian Stochastic Approximation Schemes | This paper analyzes the convergence for a large class of Riemannian
stochastic approximation (SA) schemes, which aim at tackling stochastic
optimization problems. In particular, the recursions we study use either the
exponential map of the considered manifold (geodesic schemes) or more general
retraction functions (retraction schemes) used as a proxy for the exponential
map. Such approximations are of great interest since they are low complexity
alternatives to geodesic schemes. Under the assumption that the mean field of
the SA is correlated with the gradient of a smooth Lyapunov function (possibly
non-convex), we show that the above Riemannian SA schemes find an
${\mathcal{O}}(b_\infty + \log n / \sqrt{n})$-stationary point (in expectation)
within ${\mathcal{O}}(n)$ iterations, where $b_\infty \geq 0$ is the asymptotic
bias. Compared to previous works, the conditions we derive are considerably
milder. First, all our analysis are global as we do not assume iterates to be
a-priori bounded. Second, we study biased SA schemes. To be more specific, we
consider the case where the mean-field function can only be estimated up to a
small bias, and/or the case in which the samples are drawn from a controlled
Markov chain. Third, the conditions on retractions required to ensure
convergence of the related SA schemes are weak and hold for well-known
examples. We illustrate our results on three machine learning problems.
| stat |
Efficient Variational Bayesian Structure Learning of Dynamic Graphical
Models | Estimating time-varying graphical models are of paramount importance in
various social, financial, biological, and engineering systems, since the
evolution of such networks can be utilized for example to spot trends, detect
anomalies, predict vulnerability, and evaluate the impact of interventions.
Existing methods require extensive tuning of parameters that control the graph
sparsity and temporal smoothness. Furthermore, these methods are
computationally burdensome with time complexity O(NP^3) for P variables and N
time points. As a remedy, we propose a low-complexity tuning-free Bayesian
approach, named BADGE. Specifically, we impose temporally-dependent
spike-and-slab priors on the graphs such that they are sparse and varying
smoothly across time. A variational inference algorithm is then derived to
learn the graph structures from the data automatically. Owning to the
pseudo-likelihood and the mean-field approximation, the time complexity of
BADGE is only O(NP^2). Additionally, by identifying the frequency-domain
resemblance to the time-varying graphical models, we show that BADGE can be
extended to learning frequency-varying inverse spectral density matrices, and
yields graphical models for multivariate stationary time series. Numerical
results on both synthetic and real data show that that BADGE can better recover
the underlying true graphs, while being more efficient than the existing
methods, especially for high-dimensional cases.
| stat |
Bayesian Combinatorial Multi-Study Factor Analysis | Analyzing multiple studies allows leveraging data from a range of sources and
populations, but until recently, there have been limited methodologies to
approach the joint unsupervised analysis of multiple high-dimensional studies.
A recent method, Bayesian Multi-Study Factor Analysis (BMSFA), identifies
latent factors common to all studies, as well as latent factors specific to
individual studies. However, BMSFA does not allow for partially shared factors,
i.e. latent factors shared by more than one but less than all studies. We
extend BMSFA by introducing a new method, Tetris, for Bayesian combinatorial
multi-study factor analysis, which identifies latent factors that can be shared
by any combination of studies. We model the subsets of studies that share
latent factors with an Indian Buffet Process. We test our method with an
extensive range of simulations, and showcase its utility not only in dimension
reduction but also in covariance estimation. Finally, we apply Tetris to
high-dimensional gene expression datasets to identify patterns in breast cancer
gene expression, both within and across known classes defined by germline
mutations.
| stat |
Modeling high-dimensional dependence among astronomical data | Fixing the relationship of a set of experimental quantities is a fundamental
issue in many scientific disciplines. In the 2D case, the classical approach is
to compute the linear correlation coefficient from a scatterplot. This method,
however, implicitly assumes a linear relationship between the variables. Such
an assumption is not always correct. With the use of the partial correlation
coefficients, an extension to the multidimensional case is possible. However,
the problem of the assumed mutual linear relationship of the variables remains.
A relatively recent approach that makes it possible to avoid this problem is
the modeling of the joint probability density function (PDF) of the data with
copulas. These are functions that contain all the information on the
relationship between two random variables. Although in principle this approach
also can work with multidimensional data, theoretical as well computational
difficulties often limit its use to the 2D case. In this paper, we consider an
approach based on so-called vine copulas, which overcomes this limitation and
at the same time is amenable to a theoretical treatment and feasible from the
computational point of view. We applied this method to published data on the
near-IR and far-IR luminosities and atomic and molecular masses of the Herschel
reference sample, a volume-limited sample in the nearby Universe. We determined
the relationship of the luminosities and gas masses and show that the far-IR
luminosity can be considered as the key parameter relating the other three
quantities. Once removed from the 4D relation, the residual relation among the
latter is negligible. This may be interpreted as the correlation between the
gas masses and near-IR luminosity being driven by the far-IR luminosity, likely
by the star formation activity of the galaxy.
| stat |
Volterra bootstrap: Resampling higher-order statistics for strictly
stationary univariate time series | We are concerned with nonparametric hypothesis testing of time series
functionals. It is known that the popular autoregressive sieve bootstrap is, in
general, not valid for statistics whose (asymptotic) distribution depends on
moments of order higher than two, irrespective of whether the data come from a
linear time series or a nonlinear one. Inspired by nonlinear system theory we
circumvent this non-validity by introducing a higher-order bootstrap scheme
based on the Volterra series representation of the process. In order to
estimate coefficients of such a representation efficiently, we rely on the
alternative formulation of Volterra operators in reproducing kernel Hilbert
space. We perform polynomial kernel regression which scales linearly with the
input dimensionality and is independent of the degree of nonlinearity. We
illustrate the applicability of the suggested Volterra-representation-based
bootstrap procedure in a simulation study where we consider strictly stationary
linear and nonlinear processes.
| stat |
Nonconvex Rectangular Matrix Completion via Gradient Descent without
$\ell_{2,\infty}$ Regularization | The analysis of nonconvex matrix completion has recently attracted much
attention in the community of machine learning thanks to its computational
convenience. Existing analysis on this problem, however, usually relies on
$\ell_{2,\infty}$ projection or regularization that involves unknown model
parameters, although they are observed to be unnecessary in numerical
simulations, see, e.g., Zheng and Lafferty [2016]. In this paper, we extend the
analysis of the vanilla gradient descent for positive semidefinite matrix
completion proposed in Ma et al. [2017] to the rectangular case, and more
significantly, improve the required sampling rate from
$O(\operatorname{poly}(\kappa)\mu^3 r^3 \log^3 n/n )$ to $O(\mu^2 r^2
\kappa^{14} \log n/n )$. Our technical ideas and contributions are potentially
useful in improving the leave-one-out analysis in other related problems.
| stat |
Reference and Probability-Matching Priors for the Parameters of a
Univariate Student $t$-Distribution | In this paper reference and probability-matching priors are derived for the
univariate Student $t$-distribution. These priors generally lead to procedures
with properties frequentists can relate to while still retaining Bayes
validity. The priors are tested by performing simulation studies. The focus is
on the relative mean squared error from the posterior median ($MSE(\nu)/\nu$)
and on the frequentist coverage of the 95\% credibility intervals for a sample
size of $n=30$. Average interval lengths of the credibility intervals as well
as the modes of the interval lengths based on 2000 simulations are also
considered. The performance of the priors are also tested on real data, namely
daily logarithmic returns of IBM stocks.
| stat |
Offline Recommender Learning Meets Unsupervised Domain Adaptation | We study the offline recommender learning problem in the presence of
selection bias in rating feedback. A current promising solution to address the
bias is to use the propensity score. However, the performance of the existing
propensity-based methods can significantly suffer from propensity estimation
bias. To solve the problem, we formulate the recommendation with selection bias
as unsupervised domain adaptation and derive a propensity-independent
generalization error bound. We further propose a novel algorithm that minimizes
the bound via adversarial learning. Our theory and algorithm do not depend on
propensity scores, and thus can result in a well-performing rating predictor
without requiring the true propensity information. Empirical evaluation
demonstrates the effectiveness and real-world applicability of the proposed
approach.
| stat |
Estimate of the Neural Network Dimension using Algebraic Topology and
Lie Theory | In this paper we present an approach to determine the smallest possible
number of neurons in a layer of a neural network in such a way that the
topology of the input space can be learned sufficiently well. We introduce a
general procedure based on persistent homology to investigate topological
invariants of the manifold on which we suspect the data set. We specify the
required dimensions precisely, assuming that there is a smooth manifold on or
near which the data are located. Furthermore, we require that this space is
connected and has a commutative group structure in the mathematical sense.
These assumptions allow us to derive a decomposition of the underlying space
whose topology is well known. We use the representatives of the $k$-dimensional
homology groups from the persistence landscape to determine an integer
dimension for this decomposition. This number is the dimension of the embedding
that is capable of capturing the topology of the data manifold. We derive the
theory and validate it experimentally on toy data sets.
| stat |
Data-driven regularization of Wasserstein barycenters with an
application to multivariate density registration | We present a framework to simultaneously align and smooth data in the form of
multiple point clouds sampled from unknown densities with support in a
d-dimensional Euclidean space. This work is motivated by applications in
bioinformatics where researchers aim to automatically homogenize large datasets
to compare and analyze characteristics within a same cell population.
Inconveniently, the information acquired is most certainly noisy due to
mis-alignment caused by technical variations of the environment. To overcome
this problem, we propose to register multiple point clouds by using the notion
of regularized barycenters (or Fr\'{e}chet mean) of a set of probability
measures with respect to the Wasserstein metric. A first approach consists in
penalizing a Wasserstein barycenter with a convex functional as recently
proposed in Bigot and al. (2018). A second strategy is to transform the
Wasserstein metric itself into an entropy regularized transportation cost
between probability measures as introduced in Cuturi (2013). The main
contribution of this work is to propose data-driven choices for the
regularization parameters involved in each approach using the
Goldenshluger-Lepski's principle. Simulated data sampled from Gaussian mixtures
are used to illustrate each method, and an application to the analysis of flow
cytometry data is finally proposed. This way of choosing of the regularization
parameter for the Sinkhorn barycenter is also analyzed through the prism of an
oracle inequality that relates the error made by such data-driven estimators to
the one of an ideal estimator.
| stat |
The Role of Governmental Weapons Procurements in Forecasting Monthly
Fatalities in Intrastate Conflicts: A Semiparametric Hierarchical Hurdle
Model | Accurate and interpretable forecasting models predicting spatially and
temporally fine-grained changes in the numbers of intrastate conflict
casualties are of crucial importance for policymakers and international
non-governmental organisations (NGOs). Using a count data approach, we propose
a hierarchical hurdle regression model to address the corresponding prediction
challenge at the monthly PRIO-grid level. More precisely, we model the
intensity of local armed conflict at a specific point in time as a three-stage
process. Stages one and two of our approach estimate whether we will observe
any casualties at the country- and grid-cell-level, respectively, while stage
three applies a regression model for truncated data to predict the number of
such fatalities conditional upon the previous two stages. Within this modelling
framework, we focus on the role of governmental arms imports as a processual
factor allowing governments to intensify or deter from fighting. We further
argue that a grid cell's geographic remoteness is bound to moderate the effects
of these military buildups. Out-of-sample predictions corroborate the
effectiveness of our parsimonious and theory-driven model, which enables full
transparency combined with accuracy in the forecasting process.
| stat |
Statistical characterization and time-series modeling of seismic noise | Developing statistical models for seismic noise is an exercise of high value
in seismic data analysis since these models play a critical role in detecting
the onset of seismic events. A majority of these models are usually built on
certain critical assumptions, namely, stationarity, linearity, and Gaussianity.
Despite their criticality, very little reported literature exists on validating
these assumptions on real seismic data. The objectives of this work are (i) to
critically study these long-held assumptions and (ii) to propose a systematic
procedure for developing appropriate time-series models. A rigorous statistical
analysis reveals that these standard assumptions do not hold for most of the
data sets under study; rather they exhibit additional special features such as
heteroskedasticity and integrating effects. Resting on these novel discoveries,
ARIMA-GARCH models are developed for seismic noise. Studies are carried out on
$185$ real-time data sets over different time intervals to study the daily and
seasonal variations in noise characteristics and model structure. Nearly all
datasets tested positive for first-order non-stationarity, heteroskedasticity,
and Gaussianity, while $19\%$ tested negative for linearity. Analysis of the
structural uniformity of the developed models with respect to daily and
seasonal variations is also presented.
| stat |
Randomized spectral co-clustering for large-scale directed networks | Directed networks are generally used to represent asymmetric relationships
among units. Co-clustering aims to cluster the senders and receivers of
directed networks simultaneously. In particular, the well-known spectral
clustering algorithm could be modified as the spectral co-clustering to
co-cluster directed networks. However, large-scale networks pose computational
challenge to it. In this paper, we leverage randomized sketching techniques to
accelerate the spectral co-clustering algorithms in order to co-cluster
large-scale directed networks more efficiently. Specifically, we derive two
series of randomized spectral co-clustering algorithms, one is
random-projection-based and the other is random-sampling-based. Theoretically,
we analyze the resulting algorithms under two generative models\textendash the
\emph{stochastic co-block model} and the \emph{degree corrected stochastic
co-block model}. The approximation error rates and misclustering error rates of
proposed two randomized spectral co-clustering algorithms are established,
which indicate better bounds than the state-of-the-art results of co-clustering
literature. Numerically, we conduct simulations to support our theoretical
results and test the efficiency of the algorithms on real networks with up to
tens of millions of nodes.
| stat |
Semiparametric Inference For Causal Effects In Graphical Models With
Hidden Variables | The last decade witnessed the development of algorithms that completely solve
the identifiability problem for causal effects in hidden variable causal models
associated with directed acyclic graphs. However, much of this machinery
remains underutilized in practice owing to the complexity of estimating
identifying functionals yielded by these algorithms. In this paper, we provide
simple graphical criteria and semiparametric estimators that bridge the gap
between identification and estimation for causal effects involving a single
treatment and a single outcome. First, we provide influence function based
doubly robust estimators that cover a significant subset of hidden variable
causal models where the effect is identifiable. We further characterize an
important subset of this class for which we demonstrate how to derive the
estimator with the lowest asymptotic variance, i.e., one that achieves the
semiparametric efficiency bound. Finally, we provide semiparametric estimators
for any single treatment causal effect parameter identified via the
aforementioned algorithms. The resulting estimators resemble influence function
based estimators that are sequentially reweighted, and exhibit a partial double
robustness property, provided the parts of the likelihood corresponding to a
set of weight models are correctly specified. Our methods are easy to implement
and we demonstrate their utility through simulations.
| stat |
More robust estimation of sample average treatment effects using Kernel
Optimal Matching in an observational study of spine surgical interventions | Inverse probability of treatment weighting (IPTW), which has been used to
estimate sample average treatment effects (SATE) using observational data,
tenuously relies on the positivity assumption and the correct specification of
the treatment assignment model, both of which are problematic assumptions in
many observational studies. Various methods have been proposed to overcome
these challenges, including truncation, covariate-balancing propensity scores,
and stable balancing weights. Motivated by an observational study in spine
surgery, in which positivity is violated and the true treatment assignment
model is unknown, we present the use of optimal balancing by Kernel Optimal
Matching (KOM) to estimate SATE. By uniformly controlling the conditional mean
squared error of a weighted estimator over a class of models, KOM
simultaneously mitigates issues of possible misspecification of the treatment
assignment model and is able to handle practical violations of the positivity
assumption, as shown in our simulation study. Using data from a clinical
registry, we apply KOM to compare two spine surgical interventions and
demonstrate how the result matches the conclusions of clinical trials that IPTW
estimates spuriously refute.
| stat |
A Sketch of Some Stochastic Models and Analysis Methods for Fiber Bundle
Failure under Increasing Tensile Load | Fiber bundle models (FBM's) have been used to model the failure of fibrous
composites as load-sharing systems since the 1960's when Rosen (1964 and 1965)
conducted some remarkable experiments on unidirectional fibrous composites.
These experiments gave seminal insights into their failure under increasing
tensile load. However, over the last thirty years FBM's have been used to model
catastrophic failure in other situations by the physical science community and
others. The purpose of this paper is to sketch some research on load-sharing
models and statistical analysis methods that have been overlooked by this
community. These are illustrated by summarizing the findings regarding Rosen's
Specimen A experiments and presenting the necessary results needed for this.
Related research about the bundle breaking strength distribution and the joint
distribution (the Gibbs measure) regarding the state (failed or unfailed) of
the bundle components at a given load per component, $s$, is also given.
| stat |
Permutation inference methods for multivariate meta-analysis | Multivariate meta-analysis is gaining prominence in evidence synthesis
research because it enables simultaneous synthesis of multiple correlated
outcome data, and random-effects models have generally been used for addressing
between-studies heterogeneities. However, coverage probabilities of confidence
regions or intervals for standard inference methods for random-effects models
(e.g., restricted maximum likelihood estimation) cannot retain their nominal
confidence levels in general, especially when the number of synthesized studies
is small because their validities depend on large sample approximations. In
this article, we provide permutation-based inference methods that enable exact
joint inferences for average outcome measures without large sample
approximations. We also provide accurate marginal inference methods under
general settings of multivariate meta-analyses. We propose effective approaches
for permutation inferences using optimal weighting based on the efficient score
statistic. The effectiveness of the proposed methods is illustrated via
applications to bivariate meta-analyses of diagnostic accuracy studies for
airway eosinophilia in asthma and a network meta-analysis for antihypertensive
drugs on incident diabetes, as well as through simulation experiments. In
numerical evaluations performed via simulations, our methods generally provided
accurate confidence regions or intervals under a broad range of settings,
whereas the current standard inference methods exhibited serious undercoverage
properties.
| stat |
Toward a principled Bayesian workflow in cognitive science | Experiments in research on memory, language, and in other areas of cognitive
science are increasingly being analyzed using Bayesian methods. This has been
facilitated by the development of probabilistic programming languages such as
Stan, and easily accessible front-end packages such as brms. The utility of
Bayesian methods, however, ultimately depends on the relevance of the Bayesian
model, in particular whether or not it accurately captures the structure of the
data and the data analyst's domain expertise. Even with powerful software, the
analyst is responsible for verifying the utility of their model. To demonstrate
this point, we introduce a principled Bayesian workflow (Betancourt, 2018) to
cognitive science. Using a concrete working example, we describe basic
questions one should ask about the model: prior predictive checks,
computational faithfulness, model sensitivity, and posterior predictive checks.
The running example for demonstrating the workflow is data on reading times
with a linguistic manipulation of object versus subject relative clause
sentences. This principled Bayesian workflow also demonstrates how to use
domain knowledge to inform prior distributions. It provides guidelines and
checks for valid data analysis, avoiding overfitting complex models to noise,
and capturing relevant data structure in a probabilistic model. Given the
increasing use of Bayesian methods, we aim to discuss how these methods can be
properly employed to obtain robust answers to scientific questions. All data
and code accompanying this paper are available from https://osf.io/b2vx9/.
| stat |
Scalable Structure Learning of Continuous-Time Bayesian Networks from
Incomplete Data | Continuous-time Bayesian Networks (CTBNs) represent a compact yet powerful
framework for understanding multivariate time-series data. Given complete data,
parameters and structure can be estimated efficiently in closed-form. However,
if data is incomplete, the latent states of the CTBN have to be estimated by
laboriously simulating the intractable dynamics of the assumed CTBN. This is a
problem, especially for structure learning tasks, where this has to be done for
each element of a super-exponentially growing set of possible structures. In
order to circumvent this notorious bottleneck, we develop a novel
gradient-based approach to structure learning. Instead of sampling and scoring
all possible structures individually, we assume the generator of the CTBN to be
composed as a mixture of generators stemming from different structures. In this
framework, structure learning can be performed via a gradient-based
optimization of mixture weights. We combine this approach with a new
variational method that allows for a closed-form calculation of this mixture
marginal likelihood. We show the scalability of our method by learning
structures of previously inaccessible sizes from synthetic and real-world data.
| stat |
A User Study of Perceived Carbon Footprint | We propose a statistical model to understand people's perception of their
carbon footprint. Driven by the observation that few people think of CO2 impact
in absolute terms, we design a system to probe people's perception from simple
pairwise comparisons of the relative carbon footprint of their actions. The
formulation of the model enables us to take an active-learning approach to
selecting the pairs of actions that are maximally informative about the model
parameters. We define a set of 18 actions and collect a dataset of 2183
comparisons from 176 users on a university campus. The early results reveal
promising directions to improve climate communication and enhance climate
mitigation.
| stat |
Implicitly Adaptive Importance Sampling | Adaptive importance sampling is a class of techniques for finding good
proposal distributions for importance sampling. Often the proposal
distributions are standard probability distributions whose parameters are
adapted based on the mismatch between the current proposal and a target
distribution. In this work, we present an implicit adaptive importance sampling
method that applies to complicated distributions which are not available in
closed form. The method iteratively matches the moments of a set of Monte Carlo
draws to weighted moments based on importance weights. We apply the method to
Bayesian leave-one-out cross-validation and show that it performs better than
many existing parametric adaptive importance sampling methods while being
computationally inexpensive.
| stat |
A single-index model with a surface-link for optimizing individualized
dose rules | This paper focuses on the problem of modeling and estimating interaction
effects between covariates and a continuous treatment variable on an outcome,
using a single-index regression approach. The primary motivation is to estimate
an optimal individualized dose rule in an observational study. To model
possibly nonlinear interaction effects between patients' covariates and a
continuous treatment variable, we employ a two-dimensional penalized spline
regression on an index-treatment domain, where the index is defined as a linear
projection of the covariates. The method is illustrated using two applications
as well as simulation experiments. A unique contribution of this work is in the
parsimonious (single-index) parametrization specifically defined for the
interaction effect term.
| stat |
Intervention in undirected Ising graphs and the partition function | Undirected graphical models have many applications in such areas as machine
learning, image processing, and, recently, psychology. Psychopathology in
particular has received a lot of attention, where symptoms of disorders are
assumed to influence each other. One of the most relevant questions practically
is on which symptom (node) to intervene to have the most impact. Interventions
in undirected graphical models is equal to conditioning, and so we have
available the machinery with the Ising model to determine the best strategy to
intervene. In order to perform such calculations the partition function is
required, which is computationally difficult. Here we use a Curie-Weiss
approach to approximate the partition function in applications of
interventions. We show that when the connection weights in the graph are equal
within each clique then we obtain exactly the correct partition function. And
if the weights vary according to a sub-Gaussian distribution, then the
approximation is exponentially close to the correct one. We confirm these
results with simulations.
| stat |
Modelling arterial travel time distribution using copulas | The estimation of travel time distribution (TTD) is critical for reliable
route guidance and provides theoretical bases and technical support for
advanced traffic management and control. The state-of-the art procedure for
estimating arterial TTD commonly assumes that the path travel time follows a
certain distribution without considering segment correlation. However, this
approach is usually unrealistic as travel times on successive segments may be
dependent. In this study, copula functions are used to model arterial TTD as
copulas are able to incorporate for segment correlation. First, segment
correlation is empirically investigated using day-to-day GPS data provided by
BMW Group for one major urban arterial in Munich, Germany. Segment TTDs are
estimated using a finite Gaussian Mixture Model (GMM). Next, several copula
models are introduced, namely Gaussian, Student-t, Clayton, and Gumbel, to
model the dependent structure between segment TTDs. The parameters of each
copula model are obtained by Maximum Log Likelihood Estimation. Then, path TTDs
comprised of consecutive segment TTDs are estimated based on the copula models.
The scalability of the model is evaluated by investigating the performance for
an increasing number of aggregated links. The best fitting copula is determined
in terms of goodness-of-fit test. The results demonstrate the advantage of the
proposed copula model for an increasing number of aggregated segments, compared
to the convolution without incorporating segment correlations.
| stat |
SAFFRON: an adaptive algorithm for online control of the false discovery
rate | In the online false discovery rate (FDR) problem, one observes a possibly
infinite sequence of $p$-values $P_1,P_2,\dots$, each testing a different null
hypothesis, and an algorithm must pick a sequence of rejection thresholds
$\alpha_1,\alpha_2,\dots$ in an online fashion, effectively rejecting the
$k$-th null hypothesis whenever $P_k \leq \alpha_k$. Importantly, $\alpha_k$
must be a function of the past, and cannot depend on $P_k$ or any of the later
unseen $p$-values, and must be chosen to guarantee that for any time $t$, the
FDR up to time $t$ is less than some pre-determined quantity $\alpha \in
(0,1)$. In this work, we present a powerful new framework for online FDR
control that we refer to as SAFFRON. Like older alpha-investing (AI)
algorithms, SAFFRON starts off with an error budget, called alpha-wealth, that
it intelligently allocates to different tests over time, earning back some
wealth on making a new discovery. However, unlike older methods, SAFFRON's
threshold sequence is based on a novel estimate of the alpha fraction that it
allocates to true null hypotheses. In the offline setting, algorithms that
employ an estimate of the proportion of true nulls are called adaptive methods,
and SAFFRON can be seen as an online analogue of the famous offline Storey-BH
adaptive procedure. Just as Storey-BH is typically more powerful than the
Benjamini-Hochberg (BH) procedure under independence, we demonstrate that
SAFFRON is also more powerful than its non-adaptive counterparts, such as LORD
and other generalized alpha-investing algorithms. Further, a monotone version
of the original AI algorithm is recovered as a special case of SAFFRON, that is
often more stable and powerful than the original. Lastly, the derivation of
SAFFRON provides a novel template for deriving new online FDR rules.
| stat |
A discriminative approach for finding and characterizing positivity
violations using decision trees | The assumption of positivity in causal inference (also known as common
support and co-variate overlap) is necessary to obtain valid causal estimates.
Therefore, confirming it holds in a given dataset is an important first step of
any causal analysis. Most common methods to date are insufficient for
discovering non-positivity, as they do not scale for modern high-dimensional
covariate spaces, or they cannot pinpoint the subpopulation violating
positivity. To overcome these issues, we suggest to harness decision trees for
detecting violations. By dividing the covariate space into mutually exclusive
regions, each with maximized homogeneity of treatment groups, decision trees
can be used to automatically detect subspaces violating positivity. By
augmenting the method with an additional random forest model, we can quantify
the robustness of the violation within each subspace. This solution is scalable
and provides an interpretable characterization of the subspaces in which
violations occur. We provide a visualization of the stratification rules that
define each subpopulation, combined with the severity of positivity violation
within it. We also provide an interactive version of the visualization that
allows a deeper dive into the properties of each subspace.
| stat |
Comparison of Multi-response Estimation Methods | Prediction performance does not always reflect the estimation behaviour of a
method. High error in estimation may necessarily not result in high prediction
error, but can lead to an unreliable prediction if test data lie in a slightly
different subspace than the training data. In addition, high estimation error
often leads to unstable estimates, and consequently, the estimated effect of
predictors on the response can not have a valid interpretation. Many research
fields show more interest in the effect of predictor variables than actual
prediction performance. This study compares some newly-developed (envelope) and
well-established (PCR, PLS) estimation methods using simulated data with
specifically designed properties such as Multicollinearity in the predictor
variables, the correlation between multiple responses and the position of
principal components corresponding to predictors that are relevant for the
response. This study aims to give some insights into these methods and help the
researchers to understand and use them for further study. Here we have, not
surprisingly, found that no single method is superior to others, but each has
its strength for some specific nature of data. In addition, the newly developed
envelope method has shown impressive results in finding relevant information
from data using significantly fewer components than the other methods.
| stat |
Semi-supervised voice conversion with amortized variational inference | In this work we introduce a semi-supervised approach to the voice conversion
problem, in which speech from a source speaker is converted into speech of a
target speaker. The proposed method makes use of both parallel and non-parallel
utterances from the source and target simultaneously during training. This
approach can be used to extend existing parallel data voice conversion systems
such that they can be trained with semi-supervision. We show that incorporating
semi-supervision improves the voice conversion performance compared to fully
supervised training when the number of parallel utterances is limited as in
many practical applications. Additionally, we find that increasing the number
non-parallel utterances used in training continues to improve performance when
the amount of parallel training data is held constant.
| stat |
Predicting Depressive Symptom Severity through Individuals' Nearby
Bluetooth Devices Count Data Collected by Mobile Phones: A Preliminary
Longitudinal Study | The Bluetooth sensor embedded in mobile phones provides an unobtrusive,
continuous, and cost-efficient means to capture individuals' proximity
information, such as the nearby Bluetooth devices count (NBDC). The continuous
NBDC data can partially reflect individuals' behaviors and status, such as
social connections and interactions, working status, mobility, and social
isolation and loneliness, which were found to be significantly associated with
depression by previous survey-based studies. This paper aims to explore the
NBDC data's value in predicting depressive symptom severity as measured via the
8-item Patient Health Questionnaire (PHQ-8). The data used in this paper
included 2,886 bi-weekly PHQ-8 records collected from 316 participants
recruited from three study sites in the Netherlands, Spain, and the UK as part
of the EU RADAR-CNS study. From the NBDC data two weeks prior to each PHQ-8
score, we extracted 49 Bluetooth features, including statistical features and
nonlinear features for measuring periodicity and regularity of individuals'
life rhythms. Linear mixed-effect models were used to explore associations
between Bluetooth features and the PHQ-8 score. We then applied hierarchical
Bayesian linear regression models to predict the PHQ-8 score from the extracted
Bluetooth features. A number of significant associations were found between
Bluetooth features and depressive symptom severity. Compared with commonly used
machine learning models, the proposed hierarchical Bayesian linear regression
model achieved the best prediction metrics, R2= 0.526, and root mean squared
error (RMSE) of 3.891. Bluetooth features can explain an extra 18.8% of the
variance in the PHQ-8 score relative to the baseline model without Bluetooth
features (R2=0.338, RMSE = 4.547).
| stat |
The More Data, the Better? Demystifying Deletion-Based Methods in Linear
Regression with Missing Data | We compare two deletion-based methods for dealing with the problem of missing
observations in linear regression analysis. One is the complete-case analysis
(CC, or listwise deletion) that discards all incomplete observations and only
uses common samples for ordinary least-squares estimation. The other is the
available-case analysis (AC, or pairwise deletion) that utilizes all available
data to estimate the covariance matrices and applies these matrices to
construct the normal equation. We show that the estimates from both methods are
asymptotically unbiased and further compare their asymptotic variances in some
typical situations. Surprisingly, using more data (i.e., AC) does not
necessarily lead to better asymptotic efficiency in many scenarios. Missing
patterns, covariance structure and true regression coefficient values all play
a role in determining which is better. We further conduct simulation studies to
corroborate the findings and demystify what has been missed or misinterpreted
in the literature. Some detailed proofs and simulation results are available in
the online supplemental materials.
| stat |
Mapping forest age using National Forest Inventory, airborne laser
scanning, and Sentinel-2 data | The age of forest stands is critical information for many aspects of forest
management and conservation but area-wide information about forest stand age
often does not exist. In this study, we developed regression models for
large-scale area-wide prediction of age in Norwegian forests. For model
development we used more than 4800 plots of the Norwegian National Forest
Inventory (NFI) distributed over Norway between 58{\deg} and 65{\deg} northern
latitude in a 181,773 km2 study area. Predictor variables were based on
airborne laser scanning (ALS), Sentinel-2, and existing public map data. We
performed model validation on an independent data set consisting of 63 spruce
stands with known age. The best modelling strategy was to fit independent
linear regression models to each observed site index (SI) level and using a SI
prediction map in the application of the models. The most important predictor
variable was an upper percentile of the ALS heights, and
root-mean-squared-errors (RMSE) ranged between 3 and 31 years (6% to 26%) for
SI-specific models, and 21 years (25%) on average. Mean deviance (MD) ranged
between -1 and 3 years. The models improved with increasing SI and the RMSE
were largest for low SI stands older than 100 years. Using a mapped SI, which
is required for practical applications, RMSE and MD on plot-level ranged from
19 to 56 years (29% to 53%), and 5 to 37 years (5% to 31%), respectively. For
the validation stands, the RMSE and MD were 12 (22%) and 2 years (3%). Tree
height estimated from airborne laser scanning and predicted site index were the
most important variables in the models describing age. Overall, we obtained
good results, especially for stands with high SI, that could be considered for
practical applications but see considerable potential for improvements, if
better SI maps were available.
| stat |
Learning Optimal Classification Trees: Strong Max-Flow Formulations | We consider the problem of learning optimal binary classification trees.
Literature on the topic has burgeoned in recent years, motivated both by the
empirical suboptimality of heuristic approaches and the tremendous improvements
in mixed-integer programming (MIP) technology. Yet, existing approaches from
the literature do not leverage the power of MIP to its full extent. Indeed,
they rely on weak formulations, resulting in slow convergence and large
optimality gaps. To fill this gap in the literature, we propose a flow-based
MIP formulation for optimal binary classification trees that has a stronger
linear programming relaxation. Our formulation presents an attractive
decomposable structure. We exploit this structure and max-flow/min-cut duality
to derive a Benders' decomposition method, which scales to larger instances. We
conduct extensive computational experiments on standard benchmark datasets on
which we show that our proposed approaches are 50 times faster than
state-of-the art MIP-based techniques and improve out of sample performance up
to 13.8%.
| stat |
On the estimation of average treatment effects with right-censored time
to event outcome and competing risks | We are interested in the estimation of average treatment effects based on
right-censored data of an observational study. We focus on causal inference of
differences between t-year absolute event risks in a situation with competing
risks. We derive doubly robust estimation equations and implement estimators
for the nuisance parameters based on working regression models for the outcome,
the censoring and the treatment distribution conditional on auxiliary baseline
covariates. We use the functional delta method to show that our estimators are
regular asymptotically linear estimators and estimate their variances based on
estimates of their influence functions. In empirical studies we assess the
robustness of the estimators and the coverage of confidence intervals. The
methods are further illustrated using data from a Danish registry study.
| stat |
On proportional volume sampling for experimental design in general
spaces | Optimal design for linear regression is a fundamental task in statistics. For
finite design spaces, recent progress has shown that random designs drawn using
proportional volume sampling (PVS) lead to approximation guarantees for
A-optimal design. PVS strikes the balance between design nodes that jointly
fill the design space, while marginally staying in regions of high mass under
the solution of a relaxed convex version of the original problem. In this
paper, we examine some of the statistical implications of a new variant of PVS
for (possibly Bayesian) optimal design. Using point process machinery, we treat
the case of a generic Polish design space. We show that not only are the
A-optimality approximation guarantees preserved, but we obtain similar
guarantees for D-optimal design that tighten recent results. Moreover, we show
that PVS can be sampled in polynomial time. Unfortunately, in spite of its
elegance and tractability, we demonstrate on a simple example that the
practical implications of general PVS are likely limited. In the second part of
the paper, we focus on applications and investigate the use of PVS as a
subroutine for stochastic search heuristics. We demonstrate that PVS is a
robust addition to the practitioner's toolbox, especially when the regression
functions are nonstandard and the design space, while low-dimensional, has a
complicated shape (e.g., nonlinear boundaries, several connected components).
| stat |
Randomization and Fair Judgment in Law and Science | Randomization procedures are used in legal and statistical applications,
aiming to shield important decisions from spurious influences. This article
gives an intuitive introduction to randomization and examines some intended
consequences of its use related to truthful statistical inference and fair
legal judgment. This article also presents an open-code Java implementation for
a cryptographically secure, statistically reliable, transparent, traceable, and
fully auditable randomization tool.
| stat |
Center-Outward Distribution Functions, Quantiles, Ranks, and Signs in
$\mathbb{R}^d$ | Univariate concepts as quantile and distribution functions involving ranks
and signs, do not canonically extend to $\mathbb{R}^d, d\geq 2$. Palliating
that has generated an abundant literature. Chapter 1 shows that, unlike the
many definitions that have been proposed so far, the measure
transportation-based ones introduced in Chernozhukov et al. (2017) enjoy all
the properties that make univariate quantiles and ranks successful tools for
semiparametric statistical inference.
We therefore propose a new center-outward definition of multivariate
distribution and quantile functions, along with their empirical counterparts,
for which we obtain a Glivenko-Cantelli result. Our approach is geometric and,
contrary to the Monge-Kantorovich one in Chernozhukov et al. (2017), does not
require any moment assumptions. The resulting ranks and signs are strictly
distribution-free, and maximal invariant under the action of a data-driven
class of (order-preserving) transformations generating the family of absolutely
continuous distributions; that property is the theoretical foundation of the
semiparametric efficiency preservation property of ranks. The corresponding
quantiles are equivariant under the same transformations.
The empirical proposed distribution functions are defined at observed values
only. A continuous extension to the entire $\mathbb{R}^d$, yielding continuous
empirical quantile contours while preserving the monotonicity and
Glivenko-Cantelli features is desirable. Such extension requires solving a
nontrivial problem of smooth interpolation under cyclical monotonicity
constraints. A complete solution of that problem is given in Chapter 2; we show
that the resulting distribution and quantile functions are Lipschitz, and
provide a sharp lower bound for the Lipschitz constants. A numerical study of
empirical center-outward quantile contours and their consistency is conducted.
| stat |
ABM: an automatic supervised feature engineering method for loss based
models based on group and fused lasso | A vital problem in solving classification or regression problem is to apply
feature engineering and variable selection on data before fed into models.One
of a most popular feature engineering method is to discretisize continous
variable with some cutting points,which is refered to as bining processing.Good
cutting points are important for improving model's ability, because wonderful
bining may ignore some noisy variance in continous variable range and keep
useful leveled information with good ordered encodings.However, to our best
knowledge a majority of cutting point selection is done via researchers domain
knownledge or some naive methods like equal-width cutting or equal-frequency
cutting.In this paper we propose an end-to-end supervised cutting point
selection method based on group and fused lasso along with the automatically
variable selection effect.We name our method \textbf{ABM}(automatic bining
machine). We firstly cut each variable range into fine grid bins and train
model with our group and group fused lasso regularization on each successive
bins.It is a method that integrates feature engineering,variable selection and
model training simultanously.And one more inspiring thing is that the method is
flexible such that it can be taken into a bunch of loss function based model
including deep neural networks.We have also implemented the method in R and
open the source code to other researchers.A Python version will also meet the
community in days.
| stat |
Error Bounds for Generalized Group Sparsity | In high-dimensional statistical inference, sparsity regularizations have
shown advantages in consistency and convergence rates for coefficient
estimation. We consider a generalized version of Sparse-Group Lasso which
captures both element-wise sparsity and group-wise sparsity simultaneously. We
state one universal theorem which is proved to obtain results on consistency
and convergence rates for different forms of double sparsity regularization.
The universality of the results lies in an generalization of various
convergence rates for single regularization cases such as LASSO and group LASSO
and also double regularization cases such as sparse-group LASSO. Our analysis
identifies a generalized norm of $\epsilon$-norm, which provides a dual
formulation for our double sparsity regularization.
| stat |
Adaptive Path Sampling in Metastable Posterior Distributions | The normalizing constant plays an important role in Bayesian computation, and
there is a large literature on methods for computing or approximating
normalizing constants that cannot be evaluated in closed form. When the
normalizing constant varies by orders of magnitude, methods based on importance
sampling can require many rounds of tuning. We present an improved approach
using adaptive path sampling, iteratively reducing gaps between the base and
target. Using this adaptive strategy, we develop two metastable sampling
schemes. They are automated in Stan and require little tuning. For a multimodal
posterior density, we equip simulated tempering with a continuous temperature.
For a funnel-shaped entropic barrier, we adaptively increase mass in bottleneck
regions to form an implicit divide-and-conquer. Both approaches empirically
perform better than existing methods for sampling from metastable
distributions, including higher accuracy and computation efficiency.
| stat |
Bayesian kernel machine regression-causal mediation analysis | Greater understanding of the pathways through which an environmental mixture
operates is important to design effective interventions. We present new
methodology to estimate the natural direct effect (NDE), natural indirect
effect (NIE), and controlled direct effects (CDEs) of a complex mixture
exposure on an outcome through a mediator variable. We implement Bayesian
Kernel Machine Regression (BKMR) to allow for all possible interactions and
nonlinear effects of 1) the co-exposures on the mediator, 2) the co-exposures
and mediator on the outcome, and 3) selected covariates on the mediator and/or
outcome. From the posterior predictive distributions of the mediator and
outcome, we simulate counterfactuals to obtain posterior samples, estimates,
and credible intervals of the mediation effects. Our simulation study
demonstrates that when the exposure-mediator and exposure-mediator-outcome
relationships are complex, BKMR-Causal Mediation Analysis performs better than
current mediation methods. We applied our methodology to quantify the
contribution of birth length as a mediator between in utero co-exposure to
arsenic, manganese and lead, and children's neurodevelopmental scores, in a
prospective birth cohort in Bangladesh. Among younger children, we found a
negative association between the metal mixture and neurodevelopment. We also
found evidence that birth length mediates the effect of exposure to the metal
mixture on neurodevelopment for younger children. If birth length were fixed to
its $75^{th}$ percentile value, the effect of the metal mixture on
neurodevelopment decreases, suggesting that nutritional interventions to help
increase birth length could potentially block the harmful effects of the metal
mixture on neurodevelopment.
| stat |
How to train your neural ODE: the world of Jacobian and kinetic
regularization | Training neural ODEs on large datasets has not been tractable due to the
necessity of allowing the adaptive numerical ODE solver to refine its step size
to very small values. In practice this leads to dynamics equivalent to many
hundreds or even thousands of layers. In this paper, we overcome this apparent
difficulty by introducing a theoretically-grounded combination of both optimal
transport and stability regularizations which encourage neural ODEs to prefer
simpler dynamics out of all the dynamics that solve a problem well. Simpler
dynamics lead to faster convergence and to fewer discretizations of the solver,
considerably decreasing wall-clock time without loss in performance. Our
approach allows us to train neural ODE-based generative models to the same
performance as the unregularized dynamics, with significant reductions in
training time. This brings neural ODEs closer to practical relevance in
large-scale applications.
| stat |
A threshold-free summary index for quantifying the capacity of
covariates to yield efficient treatment rules | The focus of this paper is on quantifying the capacity of covariates in
devising efficient treatment rules when data from a randomized trial are
available. Conventional one-variable-at-a-time subgroup analysis based on
statistical hypothesis testing of covariate-by-treatment interaction is
ill-suited for this purpose. The application of decision theory results in
treatment rules that compare the expected benefit of treatment given the
patient's covariates against a treatment threshold. However, determining
treatment threshold is often context-specific, and any given threshold might
seem arbitrary at the reporting stages of a clinical trial. We propose a
threshold-free metric that quantifies the capacity of a set of covariates
towards finding individuals who will benefit the most from treatment. The
construct of the proposed metric is comparing the expected outcomes with and
without knowledge of covariates when one of a two randomly selected patients
are to be treated. We show that the resulting index can also be expressed in
terms of integrated treatment benefit as a function of covariates over the
entire range of treatment thresholds. We also propose a semi-parametric
estimation method suitable for out-of-sample validation and adjustment for
optimism. We use data from a clinical trial of preventive antibiotic therapy
for reducing exacerbation rate in Chronic Obstructive Pulmonary Disease to
demonstrate the calculations in a step-by-step fashion. The proposed index has
intuitive and theoretically sound interpretation and can be estimated with
relative ease for a wide class of regression models. Beyond the conceptual
developments presented in this work, various aspects of estimation and
inference for such metrics need to be pursued in future research.
| stat |
Comment on Glenn Shafer's "Testing by betting" | This note is my comment on Glenn Shafer's discussion paper "Testing by
betting", together with two online appendices comparing p-values and betting
scores.
| stat |
Review of Mathematical frameworks for Fairness in Machine Learning | A review of the main fairness definitions and fair learning methodologies
proposed in the literature over the last years is presented from a mathematical
point of view. Following our independence-based approach, we consider how to
build fair algorithms and the consequences on the degradation of their
performance compared to the possibly unfair case. This corresponds to the price
for fairness given by the criteria $\textit{statistical parity}$ or
$\textit{equality of odds}$. Novel results giving the expressions of the
optimal fair classifier and the optimal fair predictor (under a linear
regression gaussian model) in the sense of $\textit{equality of odds}$ are
presented.
| stat |
Bayesian mitigation of spatial coarsening for a fairly flexible
spatiotemporal Hawkes model | Self-exciting spatiotemporal Hawkes processes have found increasing use in
the study of large-scale public health threats ranging from gun violence and
earthquakes to wildfires and viral contagion. Whereas many such applications
feature locational uncertainty, i.e., the exact spatial positions of individual
events are unknown, most Hawkes model analyses to date have ignored spatial
coarsening present in the data. Three particular 21st century public health
crises---urban gun violence, rural wildfires and global viral spread---present
qualitatively and quantitatively varying uncertainty regimes that exhibit (a)
different collective magnitudes of spatial coarsening, (b) uniform and mixed
magnitude coarsening, (c) differently shaped uncertainty regions and---less
orthodox---(d) locational data distributed within the `wrong' effective space.
We explicitly model such uncertainties in a Bayesian manner and jointly infer
unknown locations together with all parameters of a reasonably flexible Hawkes
model, obtaining results that are practically and statistically distinct from
those obtained while ignoring spatial coarsening. This work also features two
different secondary contributions: first, to facilitate Bayesian inference of
locations and background rate parameters, we make a subtle yet crucial change
to an established kernel-based rate model; and second, to facilitate the same
Bayesian inference at scale, we develop a massively parallel implementation of
the model's log-likelihood gradient with respect to locations and thus avoid
its quadratic computational cost in the context of Hamiltonian Monte Carlo. Our
examples involve thousands of observations and allow us to demonstrate
practicality at moderate scales.
| stat |
The temporal overfitting problem with applications in wind power curve
modeling | This paper is concerned with a nonparametric regression problem in which the
independence assumption of the input variables and the residuals is no longer
valid. Using existing model selection methods, like cross validation, the
presence of temporal autocorrelation in the input variables and the error terms
leads to model overfitting. This phenomenon is referred to as temporal
overfitting, which causes loss of performance while predicting responses for a
time domain different from the training time domain. We propose a new method to
tackle the temporal overfitting problem. Our nonparametric model is partitioned
into two parts -- a time-invariant component and a time-varying component, each
of which is modeled through a Gaussian process regression. The key in our
inference is a thinning-based strategy, an idea borrowed from Markov chain
Monte Carlo sampling, to estimate the two components, respectively. Our
specific application in this paper targets the power curve modeling in wind
energy. In our numerical studies, we compare extensively our proposed method
with both existing power curve models and available ideas for handling temporal
overfitting. Our approach yields significant improvement in prediction both in
and outside the time domain covered by the training data.
| stat |
Subsets and Splits