title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Bootstrap Inference for Multiple Imputation under Uncongeniality and
Misspecification | Multiple imputation has become one of the most popular approaches for
handling missing data in statistical analyses. Part of this success is due to
Rubin's simple combination rules. These give frequentist valid inferences when
the imputation and analysis procedures are so called congenial and the complete
data analysis is valid, but otherwise may not. Roughly speaking, congeniality
corresponds to whether the imputation and analysis models make different
assumptions about the data. In practice imputation and analysis procedures are
often not congenial, such that tests may not have the correct size and
confidence interval coverage deviates from the advertised level. We examine a
number of recent proposals which combine bootstrapping with multiple
imputation, and determine which are valid under uncongeniality and model
misspecification. Imputation followed by bootstrapping generally does not
result in valid variance estimates under uncongeniality or misspecification,
whereas bootstrapping followed by imputation does. We recommend a particular
computationally efficient variant of bootstrapping followed by imputation.
| stat |
Estimating NBA players salary share according to their performance on
court: A machine learning approach | It is customary for researchers and practitioners to fit linear models in
order to predict NBA player's salary based on the players' performance on
court. On the contrary, we focus on the players salary share (with regards to
the team payroll) by first selecting the most important determinants or
statistics (years of experience in the league, games played, etc.) and then
utilise them to predict the player salaries by employing a non linear Random
Forest machine learning algorithm. We externally evaluate our salary
predictions, thus we avoid the phenomenon of over-fitting observed in most
papers. Overall, using data from three distinct periods, 2017-2019 we identify
the important factors that achieve very satisfactory salary predictions and we
draw useful conclusions.
| stat |
Reducing Exploration of Dying Arms in Mortal Bandits | Mortal bandits have proven to be extremely useful for providing news article
recommendations, running automated online advertising campaigns, and for other
applications where the set of available options changes over time. Previous
work on this problem showed how to regulate exploration of new arms when they
have recently appeared, but they do not adapt when the arms are about to
disappear. Since in most applications we can determine either exactly or
approximately when arms will disappear, we can leverage this information to
improve performance: we should not be exploring arms that are about to
disappear. We provide adaptations of algorithms, regret bounds, and experiments
for this study, showing a clear benefit from regulating greed
(exploration/exploitation) for arms that will soon disappear. We illustrate
numerical performance on the Yahoo! Front Page Today Module User Click Log
Dataset.
| stat |
A novel algorithmic approach to Bayesian Logic Regression | Logic regression was developed more than a decade ago as a tool to construct
predictors from Boolean combinations of binary covariates. It has been mainly
used to model epistatic effects in genetic association studies, which is very
appealing due to the intuitive interpretation of logic expressions to describe
the interaction between genetic variations. Nevertheless logic regression has
(partly due to computational challenges) remained less well known than other
approaches to epistatic association mapping. Here we will adapt an advanced
evolutionary algorithm called GMJMCMC (Genetically modified Mode Jumping Markov
Chain Monte Carlo) to perform Bayesian model selection in the space of logic
regression models. After describing the algorithmic details of GMJMCMC we
perform a comprehensive simulation study that illustrates its performance given
logic regression terms of various complexity. Specifically GMJMCMC is shown to
be able to identify three-way and even four-way interactions with relatively
large power, a level of complexity which has not been achieved by previous
implementations of logic regression. We apply GMJMCMC to reanalyze QTL mapping
data for Recombinant Inbred Lines in \textit{Arabidopsis thaliana} and from a
backcross population in \textit{Drosophila} where we identify several
interesting epistatic effects. The method is implemented in an R package which
is available on github.
| stat |
A spectral algorithm for robust regression with subgaussian rates | We study a new linear up to quadratic time algorithm for linear regression in
the absence of strong assumptions on the underlying distributions of samples,
and in the presence of outliers. The goal is to design a procedure which comes
with actual working code that attains the optimal sub-gaussian error bound even
though the data have only finite moments (up to $L_4$) and in the presence of
possibly adversarial outliers. A polynomial-time solution to this problem has
been recently discovered but has high runtime due to its use of Sum-of-Square
hierarchy programming. At the core of our algorithm is an adaptation of the
spectral method introduced for the mean estimation problem to the linear
regression problem. As a by-product we established a connection between the
linear regression problem and the furthest hyperplane problem. From a
stochastic point of view, in addition to the study of the classical quadratic
and multiplier processes we introduce a third empirical process that comes
naturally in the study of the statistical properties of the algorithm.
| stat |
Precision Matrix Estimation with Noisy and Missing Data | Estimating conditional dependence graphs and precision matrices are some of
the most common problems in modern statistics and machine learning. When data
are fully observed, penalized maximum likelihood-type estimators have become
standard tools for estimating graphical models under sparsity conditions.
Extensions of these methods to more complex settings where data are
contaminated with additive or multiplicative noise have been developed in
recent years. In these settings, however, the relative performance of different
methods is not well understood and algorithmic gaps still exist. In particular,
in high-dimensional settings these methods require using non-positive
semidefinite matrices as inputs, presenting novel optimization challenges. We
develop an alternating direction method of multipliers (ADMM) algorithm for
these problems, providing a feasible algorithm to estimate precision matrices
with indefinite input and potentially nonconvex penalties. We compare this
method with existing alternative solutions and empirically characterize the
tradeoffs between them. Finally, we use this method to explore the networks
among US senators estimated from voting records data.
| stat |
Optimal sequential decision making with probabilistic digital twins | Digital twins are emerging in many industries, typically consisting of
simulation models and data associated with a specific physical system. One of
the main reasons for developing a digital twin, is to enable the simulation of
possible consequences of a given action, without the need to interfere with the
physical system itself. Physical systems of interest, and the environments they
operate in, do not always behave deterministically. Moreover, information about
the system and its environment is typically incomplete or imperfect.
Probabilistic representations of systems and environments may therefore be
called for, especially to support decisions in application areas where actions
may have severe consequences.
In this paper we introduce the probabilistic digital twin (PDT). We will
start by discussing how epistemic uncertainty can be treated using measure
theory, by modelling epistemic information via $\sigma$-algebras. Based on
this, we give a formal definition of how epistemic uncertainty can be updated
in a PDT. We then study the problem of optimal sequential decision making. That
is, we consider the case where the outcome of each decision may inform the
next. Within the PDT framework, we formulate this optimization problem. We
discuss how this problem may be solved (at least in theory) via the maximum
principle method or the dynamic programming principle. However, due to the
curse of dimensionality, these methods are often not tractable in practice. To
mend this, we propose a generic approximate solution using deep reinforcement
learning together with neural networks defined on sets. We illustrate the
method on a practical problem, considering optimal information gathering for
the estimation of a failure probability.
| stat |
Designing Experiments Informed by Observational Studies | The increasing availability of passively observed data has yielded a growing
methodological interest in "data fusion." These methods involve merging data
from observational and experimental sources to draw causal conclusions -- and
they typically require a precarious tradeoff between the unknown bias in the
observational dataset and the often-large variance in the experimental dataset.
We propose an alternative approach to leveraging observational data, which
avoids this tradeoff: rather than using observational data for inference, we
use it to design a more efficient experiment. We consider the case of a
stratified experiment with a binary outcome, and suppose pilot estimates for
the stratum potential outcome variances can be obtained from the observational
study. We extend results from Zhao et al. (2019) in order to generate
confidence sets for these variances, while accounting for the possibility of
unmeasured confounding. Then, we pose the experimental design problem as one of
regret minimization, subject to the constraints imposed by our confidence sets.
We show that this problem can be converted into a convex minimization and
solved using conventional methods. Lastly, we demonstrate the practical utility
of our methods using data from the Women's Health Initiative.
| stat |
Nonparametric instrumental variable regression and quantile regression
with full independence | The problem of endogeneity in statistics and econometrics is often handled by
introducing instrumental variables (IV) which are assumed to be mean
independent of some regressors or other observables. When full independence of
IV's and observables is assumed, nonparametric IV regression models and
nonparametric demand models lead to nonlinear integral equations with unknown
integral kernels. We prove convergence rates for the mean integrated square
error of the iteratively regularized Newton method applied to these problems.
Compared to related results we derive stronger convergence results that rely on
weaker nonlinearity restrictions. We demonstrate in numerical simulations for a
nonparametric IV regression that the method produces better results than the
standard model.
| stat |
The Brier Score under Administrative Censoring: Problems and Solutions | The Brier score is commonly used for evaluating probability predictions. In
survival analysis, with right-censored observations of the event times, this
score can be weighted by the inverse probability of censoring (IPCW) to retain
its original interpretation. It is common practice to estimate the censoring
distribution with the Kaplan-Meier estimator, even though it assumes that the
censoring distribution is independent of the covariates. This paper discusses
the general impact of the censoring estimates on the Brier score and shows that
the estimation of the censoring distribution can be problematic. In particular,
when the censoring times can be identified from the covariates, the IPCW score
is no longer valid. For administratively censored data, where the potential
censoring times are known for all individuals, we propose an alternative
version of the Brier score. This administrative Brier score does not require
estimation of the censoring distribution and is valid even if the censoring
times can be identified from the covariates.
| stat |
Accounting for Smoking in Forecasting Mortality and Life Expectancy | Smoking is one of the main risk factors that has affected human mortality and
life expectancy over the past century. Smoking accounts for a large part of the
nonlinearities in the growth of life expectancy and of the geographic and sex
differences in mortality. As Bongaarts (2006) and Janssen (2018) suggested,
accounting for smoking could improve the quality of mortality forecasts due to
the predictable nature of the smoking epidemic. We propose a new Bayesian
hierarchical model to forecast life expectancy at birth for both sexes and for
69 countries with good data on smoking-related mortality. The main idea is to
convert the forecast of the non-smoking life expectancy at birth (i.e., life
expectancy at birth removing the smoking effect) into life expectancy forecast
through the use of the age-specific smoking attributable fraction (ASSAF). We
introduce a new age-cohort model for the ASSAF and a Bayesian hierarchical
model for non-smoking life expectancy at birth. The forecast performance of the
proposed method is evaluated by out-of-sample validation compared with four
other commonly used methods for life expectancy forecasting. Improvements in
forecast accuracy and model calibration based on the new method are observed.
| stat |
A new approach in model selection for ordinal target variables | This paper introduces a novel approach to assess model performance for
predictive models characterized by an ordinal target variable in order to
satisfy the lack of suitable tools in this framework. Our methodological
proposal is a new index for model assessment which satisfies mathematical
properties and can be easily computed. In order to show how our performance
indicator works, empirical evidence achieved on a toy examples and simulated
data are provided. On the basis of results at hand, we underline that our
approach discriminates better for model selection with respect to performance
indexes proposed in the literature.
| stat |
Query Complexity of k-NN based Mode Estimation | Motivated by the mode estimation problem of an unknown multivariate
probability density function, we study the problem of identifying the point
with the minimum k-th nearest neighbor distance for a given dataset of n
points. We study the case where the pairwise distances are apriori unknown, but
we have access to an oracle which we can query to get noisy information about
the distance between any pair of points. For two natural oracle models, we
design a sequential learning algorithm, based on the idea of confidence
intervals, which adaptively decides which queries to send to the oracle and is
able to correctly solve the problem with high probability. We derive
instance-dependent upper bounds on the query complexity of our proposed scheme
and also demonstrate significant improvement over the performance of other
baselines via extensive numerical evaluations.
| stat |
Assessing Data Support for the Simplifying Assumption in Bivariate
Conditional Copulas | The paper considers the problem of establishing data support for the
simplifying assumption (SA) in a bivariate conditional copula model. It is
known that SA greatly simplifies the inference for a conditional copula model,
but standard tools and methods for testing SA tend to not provide reliable
results. After splitting the observed data into training and test sets, the
method proposed will use a flexible training data Bayesian fit to define tests
based on randomization and standard asymptotic theory. Theoretical
justification for the method is provided and its performance is studied using
simulated data. The paper also discusses implementations in alternative models
of interest, e.g. Gaussian, Logistic and Quantile regressions.
| stat |
Forecasting multiple functional time series in a group structure: an
application to mortality | When modeling sub-national mortality rates, we should consider three
features: (1) how to incorporate any possible correlation among sub-populations
to potentially improve forecast accuracy through multi-population joint
modeling; (2) how to reconcile sub-national mortality forecasts so that they
aggregate adequately across various levels of a group structure; (3) among the
forecast reconciliation methods, how to combine their forecasts to achieve
improved forecast accuracy. To address these issues, we introduce an extension
of grouped univariate functional time series method. We first consider a
multivariate functional time series method to jointly forecast multiple related
series. We then evaluate the impact and benefit of using forecast combinations
among the forecast reconciliation methods. Using the Japanese regional
age-specific mortality rates, we investigate one-step-ahead to 15-step-ahead
point and interval forecast accuracies of our proposed extension and make
recommendations.
| stat |
Robust Learning Rate Selection for Stochastic Optimization via Splitting
Diagnostic | This paper proposes SplitSGD, a new dynamic learning rate schedule for
stochastic optimization. This method decreases the learning rate for better
adaptation to the local geometry of the objective function whenever a
stationary phase is detected, that is, the iterates are likely to bounce at
around a vicinity of a local minimum. The detection is performed by splitting
the single thread into two and using the inner product of the gradients from
the two threads as a measure of stationarity. Owing to this simple yet provably
valid stationarity detection, SplitSGD is easy-to-implement and essentially
does not incur additional computational cost than standard SGD. Through a
series of extensive experiments, we show that this method is appropriate for
both convex problems and training (non-convex) neural networks, with
performance compared favorably to other stochastic optimization methods.
Importantly, this method is observed to be very robust with a set of default
parameters for a wide range of problems and, moreover, yields better
generalization performance than other adaptive gradient methods such as Adam.
| stat |
Optimal Robust Linear Regression in Nearly Linear Time | We study the problem of high-dimensional robust linear regression where a
learner is given access to $n$ samples from the generative model $Y = \langle
X,w^* \rangle + \epsilon$ (with $X \in \mathbb{R}^d$ and $\epsilon$
independent), in which an $\eta$ fraction of the samples have been
adversarially corrupted. We propose estimators for this problem under two
settings: (i) $X$ is L4-L2 hypercontractive, $\mathbb{E} [XX^\top]$ has bounded
condition number and $\epsilon$ has bounded variance and (ii) $X$ is
sub-Gaussian with identity second moment and $\epsilon$ is sub-Gaussian. In
both settings, our estimators: (a) Achieve optimal sample complexities and
recovery guarantees up to log factors and (b) Run in near linear time
($\tilde{O}(nd / \eta^6)$). Prior to our work, polynomial time algorithms
achieving near optimal sample complexities were only known in the setting where
$X$ is Gaussian with identity covariance and $\epsilon$ is Gaussian, and no
linear time estimators were known for robust linear regression in any setting.
Our estimators and their analysis leverage recent developments in the
construction of faster algorithms for robust mean estimation to improve
runtimes, and refined concentration of measure arguments alongside Gaussian
rounding techniques to improve statistical sample complexities.
| stat |
Bayesian Experimental Design for Oral Glucose Tolerance Tests (OGTT) | OGTT is a common test, frequently used to diagnose insulin resistance or
diabetes, in which a patient's blood sugar is measured at various times over
the course of a few hours. Recent developments in the study of OGTT results
have framed it as an inverse problem which has been the subject of Bayesian
inference. This is a powerful new tool for analyzing the results of an OGTT
test,and the question arises as to whether the test itself can be improved. It
is of particular interest to discover whether the times at which a patient's
glucose is measured can be changed to improve the effectiveness of the test.
The purpose of this paper is to explore the possibility of finding a better
experimental design, that is, a set of times to perform the test. We review the
theory of Bayesian experimental design and propose an estimator for the
expected utility of a design. We then study the properties of this estimator
and propose a new method for quantifying the uncertainty in comparisons between
designs. We implement this method to find a new design and the proposed design
is compared favorably to the usual testing scheme.
| stat |
Multivariate time-series modeling with generative neural networks | Generative moment matching networks (GMMNs) are introduced as dependence
models for the joint innovation distribution of multivariate time series (MTS).
Following the popular copula-GARCH approach for modeling dependent MTS data, a
framework based on a GMMN-GARCH approach is presented. First, ARMA-GARCH models
are utilized to capture the serial dependence within each univariate marginal
time series. Second, if the number of marginal time series is large, principal
component analysis (PCA) is used as a dimension-reduction step. Last, the
remaining cross-sectional dependence is modeled via a GMMN, the main
contribution of this work. GMMNs are highly flexible and easy to simulate from,
which is a major advantage over the copula-GARCH approach. Applications
involving yield curve modeling and the analysis of foreign exchange-rate
returns demonstrate the utility of the GMMN-GARCH approach, especially in terms
of producing better empirical predictive distributions and making better
probabilistic forecasts.
| stat |
Partial Correlation Graphical LASSO | Standard likelihood penalties to learn Gaussian graphical models are based on
regularising the off-diagonal entries of the precision matrix. Such methods,
and their Bayesian counterparts, are not invariant to scalar multiplication of
the variables, unless one standardises the observed data to unit sample
variances. We show that such standardisation can have a strong effect on
inference and introduce a new family of penalties based on partial
correlations. We show that the latter, as well as the maximum likelihood, $L_0$
and logarithmic penalties are scale invariant. We illustrate the use of one
such penalty, the partial correlation graphical LASSO, which sets an $L_{1}$
penalty on partial correlations. The associated optimization problem is no
longer convex, but is conditionally convex. We show via simulated examples and
in two real datasets that, besides being scale invariant, there can be
important gains in terms of inference.
| stat |
Iterative Method for Tuning Complex Simulation Code | Tuning a complex simulation code refers to the process of improving the
agreement of a code calculation with respect to a set of experimental data by
adjusting parameters implemented in the code. This process belongs to the class
of inverse problems or model calibration. For this problem, the approximated
nonlinear least squares (ANLS) method based on a Gaussian process (GP)
metamodel has been employed by some researchers. A potential drawback of the
ANLS method is that the metamodel is built only once and not updated
thereafter. To address this difficulty, we propose an iterative algorithm in
this study. In the proposed algorithm, the parameters of the simulation code
and GP metamodel are alternatively re-estimated and updated by maximum
likelihood estimation and the ANLS method. This algorithm uses both computer
and experimental data repeatedly until convergence. A study using toy-models
including inexact computer code with bias terms reveals that the proposed
algorithm performs better than the ANLS method and the
conditional-likelihood-based approach. Finally, an application to a nuclear
fusion simulation code is illustrated.
| stat |
Gaussian kernel smoothing | Image acquisition and segmentation are likely to introduce noise. Further
image processing such as image registration and parameterization can introduce
additional noise. It is thus imperative to reduce noise measurements and boost
signal. In order to increase the signal-to-noise ratio (SNR) and smoothness of
data required for the subsequent random field theory based statistical
inference, some type of smoothing is necessary. Among many image smoothing
methods, Gaussian kernel smoothing has emerged as a de facto smoothing
technique among brain imaging researchers due to its simplicity in numerical
implementation. Gaussian kernel smoothing also increases statistical
sensitivity and statistical power as well as Gausianness. Gaussian kernel
smoothing can be viewed as weighted averaging of voxel values. Then from the
central limit theorem, the weighted average should be more Gaussian.
| stat |
Alleviating Spatial Confounding in Spatial Frailty Models | Spatial confounding is how is called the confounding between fixed and
spatial random effects. It has been widely studied and it gained attention in
the past years in the spatial statistics literature, as it may generate
unexpected results in modeling. The projection-based approach, also known as
restricted models, appears as a good alternative to overcome the spatial
confounding in generalized linear mixed models. However, when the support of
fixed effects is different from the spatial effect one, this approach can no
longer be applied directly. In this work, we introduce a method to alleviate
the spatial confounding for the spatial frailty models family. This class of
models can incorporate spatially structured effects and it is usual to observe
more than one sample unit per area which means that the support of fixed and
spatial effects differs. In this case, we introduce a two folded
projection-based approach projecting the design matrix to the dimension of the
space and then projecting the random effect to the orthogonal space of the new
design matrix. To provide fast inference in our analysis we employ the
integrated nested Laplace approximation methodology. The method is illustrated
with an application with lung and bronchus cancer in California - US that
confirms that the methodology efficiency.
| stat |
Group segmentation and heterogeneity in the choice of cooking fuels in
post-earthquake Nepal | Segmenting population into subgroups with higher intergroup, but lower
intragroup, heterogeneity can be useful in enhancing the effectiveness of many
socio-economic policy interventions; yet it has received little attention in
promoting clean cooking. Here, we use PERMANOVA, a distance-based multivariate
analysis, to identify the factor that captures the highest intergroup
heterogeneity in the choice of cooking fuels. Applying this approach to the
post-earthquake data on 747,137 households from Nepal, we find that ethnicity
explains 39.12% of variation in fuel choice, followed by income (26.30%),
education (12.62%), and location (4.05%). This finding indicates that
ethnicity, rather than income or other factors, as a basis of policy
interventions may be more effective in promoting clean cooking. We also find
that, among the ethnic groups in Nepal, the most marginalized Chepang/Thami
community exhibits the lowest intragroup diversity (Shannon index = 0.101)
while Newars the highest (0.667). This information on intra-ethnic diversity in
fuel choice can have important policy implications for reducing ethnic gap in
clean cooking.
| stat |
Elastic Net Regularization Paths for All Generalized Linear Models | The lasso and elastic net are popular regularized regression models for
supervised learning. Friedman, Hastie, and Tibshirani (2010) introduced a
computationally efficient algorithm for computing the elastic net
regularization path for ordinary least squares regression, logistic regression
and multinomial logistic regression, while Simon, Friedman, Hastie, and
Tibshirani (2011) extended this work to Cox models for right-censored data. We
further extend the reach of the elastic net-regularized regression to all
generalized linear model families, Cox models with (start, stop] data and
strata, and a simplified version of the relaxed lasso. We also discuss
convenient utility functions for measuring the performance of these fitted
models.
| stat |
On Statistical Bias In Active Learning: How and When To Fix It | Active learning is a powerful tool when labelling data is expensive, but it
introduces a bias because the training data no longer follows the population
distribution. We formalize this bias and investigate the situations in which it
can be harmful and sometimes even helpful. We further introduce novel
corrective weights to remove bias when doing so is beneficial. Through this,
our work not only provides a useful mechanism that can improve the active
learning approach, but also an explanation of the empirical successes of
various existing approaches which ignore this bias. In particular, we show that
this bias can be actively helpful when training overparameterized models --
like neural networks -- with relatively little data.
| stat |
A Simulation Study Evaluating Phase I Clinical Trial Designs for
Combinational Agents | Nowadays, more and more clinical trials choose combinational agents as the
intervention to achieve better therapeutic responses. However, dose-finding for
combinational agents is much more complicated than single agent as the full
order of combination dose toxicity is unknown. Therefore, regular phase I
designs are not able to identify the maximum tolerated dose (MTD) of
combinational agents. Motivated by such needs, plenty of novel phase I clinical
trial designs for combinational agents were proposed. With so many available
designs, research that compare their performances, explore parameters' impacts,
and provide recommendations is very limited. Therefore, we conducted a
simulation study to evaluate multiple phase I designs that proposed to identify
single MTD for combinational agents under various scenarios. We also explored
influences of different design parameters. In the end, we summarized the pros
and cons of each design, and provided a general guideline in design selection.
| stat |
Infinite mixtures of multivariate normal-inverse Gaussian distributions
for clustering of skewed data | Mixtures of multivariate normal inverse Gaussian (MNIG) distributions can be
used to cluster data that exhibit features such as skewness and heavy tails.
However, for cluster analysis, using a traditional finite mixture model
framework, either the number of components needs to be known $a$-$priori$ or
needs to be estimated $a$-$posteriori$ using some model selection criterion
after deriving results for a range of possible number of components. However,
different model selection criteria can sometimes result in different number of
components yielding uncertainty. Here, an infinite mixture model framework,
also known as Dirichlet process mixture model, is proposed for the mixtures of
MNIG distributions. This Dirichlet process mixture model approach allows the
number of components to grow or decay freely from 1 to $\infty$ (in practice
from 1 to $N$) and the number of components is inferred along with the
parameter estimates in a Bayesian framework thus alleviating the need for model
selection criteria. We provide real data applications with benchmark datasets
as well as a small simulation experiment to compare with other existing models.
The proposed method provides competitive clustering results to other clustering
approaches for both simulation and real data and parameter recovery are
illustrated using simulation studies.
| stat |
Bayesian Markov Switching Tensor Regression for Time-varying Networks | We propose a new Bayesian Markov switching regression model for
multidimensional arrays (tensors) of binary time series. We assume a
zero-inflated logit regression with time-varying parameters and apply it to
multilayer temporal networks. The original contribution is threefold. First, to
avoid over-fitting we propose a parsimonious parametrization based on a
low-rank decomposition of the tensor of regression coefficients. Second, we
assume the parameters are driven by a hidden Markov chain, thus allowing for
structural changes in the network topology. We follow a Bayesian approach to
inference and provide an efficient Gibbs sampler for posterior approximation.
We apply the methodology to a real dataset of financial networks to study the
impact of several risk factors on the edge probability. Supplementary materials
for this article are available online.
| stat |
Stochastic Approximation Cut Algorithm for Inference in Modularized
Bayesian Models | Bayesian modelling enables us to accommodate complex forms of data and make a
comprehensive inference, but the effect of partial misspecification of the
model is a concern. One approach in this setting is to modularize the model,
and prevent feedback from suspect modules, using a cut model. After observing
data, this leads to the cut distribution which normally does not have a
closed-form. Previous studies have proposed algorithms to sample from this
distribution, but these algorithms have unclear theoretical convergence
properties. To address this, we propose a new algorithm called the Stochastic
Approximation Cut algorithm (SACut) as an alternative. The algorithm is divided
into two parallel chains. The main chain targets an approximation to the cut
distribution; the auxiliary chain is used to form an adaptive proposal
distribution for the main chain. We prove convergence of the samples drawn by
the proposed algorithm and present the exact limit. Although SACut is biased,
since the main chain does not target the exact cut distribution, we prove this
bias can be reduced geometrically by increasing a user-chosen tuning parameter.
In addition, parallel computing can be easily adopted for SACut, which greatly
reduces computation time.
| stat |
On generalization in moment-based domain adaptation | Domain adaptation algorithms are designed to minimize the misclassification
risk of a discriminative model for a target domain with little training data by
adapting a model from a source domain with a large amount of training data.
Standard approaches measure the adaptation discrepancy based on distance
measures between the empirical probability distributions in the source and
target domain. In this setting, we address the problem of deriving learning
bounds under practice-oriented general conditions on the underlying probability
distributions. As a result, we obtain generalization bounds for domain
adaptation based on finitely many moments and smoothness conditions.
| stat |
Dynamic Slate Recommendation with Gated Recurrent Units and Thompson
Sampling | We consider the problem of recommending relevant content to users of an
internet platform in the form of lists of items, called slates. We introduce a
variational Bayesian Recurrent Neural Net recommender system that acts on time
series of interactions between the internet platform and the user, and which
scales to real world industrial situations. The recommender system is tested
both online on real users, and on an offline dataset collected from a Norwegian
web-based marketplace, FINN.no, that is made public for research. This is one
of the first publicly available datasets which includes all the slates that are
presented to users as well as which items (if any) in the slates were clicked
on. Such a data set allows us to move beyond the common assumption that
implicitly assumes that users are considering all possible items at each
interaction. Instead we build our likelihood using the items that are actually
in the slate, and evaluate the strengths and weaknesses of both approaches
theoretically and in experiments. We also introduce a hierarchical prior for
the item parameters based on group memberships. Both item parameters and user
preferences are learned probabilistically. Furthermore, we combine our model
with bandit strategies to ensure learning, and introduce `in-slate Thompson
Sampling' which makes use of the slates to maximise explorative opportunities.
We show experimentally that explorative recommender strategies perform on par
or above their greedy counterparts. Even without making use of exploration to
learn more effectively, click rates increase simply because of improved
diversity in the recommended slates.
| stat |
Bayesian analysis of dynamic binary data: A simulation study and
application to economic index SP | It is proposed in the literature that in some complicated problems maximum
likelihood estimates (MLE) are not suitable or even do not exist. An
alternative to MLE for estimation of the parameters is the Bayesian method. The
Markov chain Monte Carlo (MCMC) simulation procedure is designed to fit
Bayesian models. Bayesian method like classical method (MLE) has advantages and
disadvantages. One of the advantages of Bayesian method over MLE method is the
ability of saving the information included in past data through the posterior
distributions of the model parameters to be used for modelling future data. In
this article we investigate the performance of Bayesian method in modelling
dynamic binary data when the data are growing over time and individuals.
| stat |
SPARQ-SGD: Event-Triggered and Compressed Communication in Decentralized
Stochastic Optimization | In this paper, we propose and analyze SPARQ-SGD, which is an event-triggered
and compressed algorithm for decentralized training of large-scale machine
learning models. Each node can locally compute a condition (event) which
triggers a communication where quantized and sparsified local model parameters
are sent. In SPARQ-SGD each node takes at least a fixed number ($H$) of local
gradient steps and then checks if the model parameters have significantly
changed compared to its last update; it communicates further compressed model
parameters only when there is a significant change, as specified by a (design)
criterion. We prove that the SPARQ-SGD converges as $O(\frac{1}{nT})$ and
$O(\frac{1}{\sqrt{nT}})$ in the strongly-convex and non-convex settings,
respectively, demonstrating that such aggressive compression, including
event-triggered communication, model sparsification and quantization does not
affect the overall convergence rate as compared to uncompressed decentralized
training; thereby theoretically yielding communication efficiency for "free".
We evaluate SPARQ-SGD over real datasets to demonstrate significant amount of
savings in communication over the state-of-the-art.
| stat |
Statistical Estimation of High-Dimensional Vector Autoregressive Models | High-dimensional vector autoregressive (VAR) models are important tools for
the analysis of multivariate time series. This paper focuses on
high-dimensional time series and on the different regularized estimation
procedures proposed for fitting sparse VAR models to such time series.
Attention is paid to the different sparsity assumptions imposed on the VAR
parameters and how these sparsity assumptions are related to the particular
consistency properties of the estimators established. A sparsity scheme for
high-dimensional VAR models is proposed which is found to be more appropriate
for the time series setting considered. Furthermore, it is shown that, under
this sparsity setting, threholding extents the consistency properties of
regularized estimators to a wide range of matrix norms. Among other things,
this enables application of the VAR parameters estimators to different
inference problems, like forecasting or estimating the second-order
characteristics of the underlying VAR process. Extensive simulations compare
the finite sample behavior of the different regularized estimators proposed
using a variety of performance criteria.
| stat |
$R^*$: A robust MCMC convergence diagnostic with uncertainty using
decision tree classifiers | Markov chain Monte Carlo (MCMC) has transformed Bayesian model inference over
the past three decades: mainly because of this, Bayesian inference is now a
workhorse of applied scientists. Under general conditions, MCMC sampling
converges asymptotically to the posterior distribution, but this provides no
guarantees about its performance in finite time. The predominant method for
monitoring convergence is to run multiple chains and monitor individual chains'
characteristics and compare these to the population as a whole: if within-chain
and between-chain summaries are comparable, then this is taken to indicate that
the chains have converged to a common stationary distribution. Here, we
introduce a new method for diagnosing convergence based on how well a machine
learning classifier model can successfully discriminate the individual chains.
We call this convergence measure $R^*$. In contrast to the predominant
$\widehat{R}$, $R^*$ is a single statistic across all parameters that indicates
lack of mixing, although individual variables' importance for this metric can
also be determined. Additionally, $R^*$ is not based on any single
characteristic of the sampling distribution; instead it uses all the
information in the chain, including that given by the joint sampling
distribution, which is currently largely overlooked by existing approaches. We
recommend calculating $R^*$ using two different machine learning classifiers -
gradient-boosted regression trees and random forests - which each work well in
models of different dimensions. Because each of these methods outputs a
classification probability, as a byproduct, we obtain uncertainty in $R^*$. The
method is straightforward to implement and could be a complementary additional
check on MCMC convergence for applied analyses.
| stat |
Identifiability of Complete Dictionary Learning | Sparse component analysis (SCA), also known as complete dictionary learning,
is the following problem: Given an input matrix $M$ and an integer $r$, find a
dictionary $D$ with $r$ columns and a matrix $B$ with $k$-sparse columns (that
is, each column of $B$ has at most $k$ non-zero entries) such that $M \approx
DB$. A key issue in SCA is identifiability, that is, characterizing the
conditions under which $D$ and $B$ are essentially unique (that is, they are
unique up to permutation and scaling of the columns of $D$ and rows of $B$).
Although SCA has been vastly investigated in the last two decades, only a few
works have tackled this issue in the deterministic scenario, and no work
provides reasonable bounds in the minimum number of samples (that is, columns
of $M$) that leads to identifiability. In this work, we provide new results in
the deterministic scenario when the data has a low-rank structure, that is,
when $D$ is (under)complete. While previous bounds feature a combinatorial term
$r \choose k$, we exhibit a sufficient condition involving
$\mathcal{O}(r^3/(r-k)^2)$ samples that yields an essentially unique
decomposition, as long as these data points are well spread among the subspaces
spanned by $r-1$ columns of $D$. We also exhibit a necessary lower bound on the
number of samples that contradicts previous results in the literature when $k$
equals $r-1$. Our bounds provide a drastic improvement compared to the state of
the art, and imply for example that for a fixed proportion of zeros (constant
and independent of $r$, e.g., 10\% of zero entries in $B$), one only requires
$\mathcal{O}(r)$ data points to guarantee identifiability.
| stat |
Jackknife Empirical Likelihood Approach for K-sample Tests | The categorical Gini correlation is an alternative measure of dependence
between a categorical and numerical variables, which characterizes the
independence of the variables. A nonparametric test for the equality of K
distributions has been developed based on the categorical Gini correlation. By
applying the jackknife empirical likelihood approach, the standard limiting
chi-square distribution with degree freedom of $K-1$ is established and is used
to determine critical value and $p$-value of the test. Simulation studies show
that the proposed method is competitive to existing methods in terms of power
of the tests in most cases. The proposed method is illustrated in an
application on a real data set.
| stat |
Un mod\`ele Bay\'esien de co-clustering de donn\'ees mixtes | We propose a MAP Bayesian approach to perform and evaluate a co-clustering of
mixed-type data tables. The proposed model infers an optimal segmentation of
all variables then performs a co-clustering by minimizing a Bayesian model
selection cost function. One advantage of this approach is that it is user
parameter-free. Another main advantage is the proposed criterion which gives an
exact measure of the model quality, measured by probability of fitting it to
the data. Continuous optimization of this criterion ensures finding better and
better models while avoiding data over-fitting. The experiments conducted on
real data show the interest of this co-clustering approach in exploratory data
analysis of large data sets.
| stat |
Decision-making with multiple correlated binary outcomes in clinical
trials | Clinical trials often evaluate multiple outcome variables to form a
comprehensive picture of the effects of a new treatment. The resulting
multidimensional insight contributes to clinically relevant and efficient
decision-making about treatment superiority. Common statistical procedures to
make these superiority decisions with multiple outcomes have two important
shortcomings however: 1) Outcome variables are often modeled individually, and
consequently fail to consider the relation between outcomes; and 2) superiority
is often defined as a relevant difference on a single, on any, or on all
outcomes(s); and lacks a compensatory mechanism that allows large positive
effects on one or multiple outcome(s) to outweigh small negative effects on
other outcomes. To address these shortcomings, this paper proposes 1) a
Bayesian model for the analysis of correlated binary outcomes based on the
multivariate Bernoulli distribution; and 2) a flexible decision criterion with
a compensatory mechanism that captures the relative importance of the outcomes.
A simulation study demonstrates that efficient and unbiased decisions can be
made while Type I error rates are properly controlled. The performance of the
framework is illustrated for 1) fixed, group sequential, and adaptive designs;
and 2) non-informative and informative prior distributions.
| stat |
Variational Orthogonal Features | Sparse stochastic variational inference allows Gaussian process models to be
applied to large datasets. The per iteration computational cost of inference
with this method is $\mathcal{O}(\tilde{N}M^2+M^3),$ where $\tilde{N}$ is the
number of points in a minibatch and $M$ is the number of `inducing features',
which determine the expressiveness of the variational family. Several recent
works have shown that for certain priors, features can be defined that remove
the $\mathcal{O}(M^3)$ cost of computing a minibatch estimate of an evidence
lower bound (ELBO). This represents a significant computational savings when
$M\gg \tilde{N}$. We present a construction of features for any stationary
prior kernel that allow for computation of an unbiased estimator to the ELBO
using $T$ Monte Carlo samples in $\mathcal{O}(\tilde{N}T+M^2T)$ and in
$\mathcal{O}(\tilde{N}T+MT)$ with an additional approximation. We analyze the
impact of this additional approximation on inference quality.
| stat |
Sparse Sliced Inverse Regression via Cholesky Matrix Penalization | We introduce a new sparse sliced inverse regression estimator called Cholesky
matrix penalization and its adaptive version for achieving sparsity in
estimating the dimensions of the central subspace. The new estimators use the
Cholesky decomposition of the covariance matrix of the covariates and include a
regularization term in the objective function to achieve sparsity in a
computationally efficient manner. We establish the theoretical values of the
tuning parameters that achieve estimation and variable selection consistency
for the central subspace. Furthermore, we propose a new projection information
criterion to select the tuning parameter for our proposed estimators and prove
that the new criterion facilitates selection consistency. The Cholesky matrix
penalization estimator inherits the strength of the Matrix Lasso and the Lasso
sliced inverse regression estimator; it has superior performance in numerical
studies and can be adapted to other sufficient dimension methods in the
literature.
| stat |
Assessment and adjustment of approximate inference algorithms using the
law of total variance | A common method for assessing validity of Bayesian sampling or approximate
inference methods makes use of simulated data replicates for parameters drawn
from the prior. Under continuity assumptions, quantiles of functions of the
simulated parameter values within corresponding posterior distributions are
uniformly distributed. Checking for uniformity when a posterior density is
approximated numerically provides a diagnostic for algorithm validity.
Furthermore, adjustments to achieve uniformity can improve the quality of
approximate inference methods. A weakness of this general approach is that it
seems difficult to extend beyond scalar functions of interest. The present
article develops an alternative to quantile-based checking and adjustment
methods which is inherently multivariate. The new approach is based on use of
the tower property of conditional expectation and the law of total variance for
relating prior and posterior expectations and covariances. For adjustment,
approximate inferences are modified so that the correct prior to posterior
relationships hold. We illustrate the method in three examples. The first uses
an auxiliary model in a likelihood-free inference problem. The second considers
corrections for variational Bayes approximations in a deep neural network
generalized linear mixed model. Our final application considers a deep neural
network surrogate for approximating Gaussian process regression predictive
inference.
| stat |
Bayesian Model Selection with Graph Structured Sparsity | We propose a general algorithmic framework for Bayesian model selection. A
spike-and-slab Laplacian prior is introduced to model the underlying structural
assumption. Using the notion of effective resistance, we derive an EM-type
algorithm with closed-form iterations to efficiently explore possible
candidates for Bayesian model selection. The deterministic nature of the
proposed algorithm makes it more scalable to large-scale and high-dimensional
data sets compared with existing stochastic search algorithms. When applied to
sparse linear regression, our framework recovers the EMVS algorithm [Rockova
and George, 2014] as a special case. We also discuss extensions of our
framework using tools from graph algebra to incorporate complex Bayesian models
such as biclustering and submatrix localization. Extensive simulation studies
and real data applications are conducted to demonstrate the superior
performance of our methods over its frequentist competitors such as $\ell_0$ or
$\ell_1$ penalization.
| stat |
Rethinking Representations in P&C Actuarial Science with Deep Neural
Networks | Insurance companies gather a growing variety of data for use in the insurance
process, but most traditional ratemaking models are not designed to support
them. In particular, many emerging data sources (text, images, sensors) may
complement traditional data to provide better insights to predict the future
losses in an insurance contract. This paper presents some of these emerging
data sources and presents a unified framework for actuaries to incorporate
these in existing ratemaking models. Our approach stems from representation
learning, whose goal is to create representations of raw data. A useful
representation will transform the original data into a dense vector space where
the ultimate predictive task is simpler to model. Our paper presents methods to
transform non-vectorial data into vectorial representations and provides
examples for actuarial science.
| stat |
Statistical learning for sensor localization in wireless networks | Indoor localization has become an important issue for wireless sensor
networks. This paper presents a zoning-based localization technique that uses
WiFi signals and works efficiently in indoor environments. The targeted area is
composed of several zones, the objective being to determine the zone of the
sensor using an observation model based on statistical learning.
| stat |
Sugar and Stops in Drivers with Insulin-Dependent Type 1 Diabetes | Diabetes is a major public health challenge worldwide. Abnormal physiology in
diabetes, particularly hypoglycemia, can cause driver impairments that affect
safe driving. While diabetes driver safety has been previously researched, few
studies link real-time physiologic changes in drivers with diabetes to
objective real-world driver safety, particularly at high-risk areas like
intersections. To address this, we investigated the role of acute physiologic
changes in drivers with type 1 diabetes mellitus (T1DM) on safe stopping at
stop intersections. 18 T1DM drivers (21-52 years, mean = 31.2 years) and 14
controls (21-55 years, mean = 33.4 years) participated in a 4-week naturalistic
driving study. At induction, each participant's vehicle was fitted with a
camera and sensor system to collect driving data. Video was processed with
computer vision algorithms detecting traffic elements. Stop intersections were
geolocated with clustering methods, state intersection databases, and manual
review. Videos showing driver stop intersection approaches were extracted and
manually reviewed to classify stopping behavior (full, rolling, and no stop)
and intersection traffic characteristics. Mixed-effects logistic regression
models determined how diabetes driver stopping safety (safe vs. unsafe stop)
was affected by 1) disease and 2) at-risk, acute physiology (hypo- and
hyperglycemia). Diabetes drivers who were acutely hyperglycemic had 2.37
increased odds of unsafe stopping (95% CI: 1.26-4.47, p = 0.008) compared to
those with normal physiology. Acute hypoglycemia did not associate with unsafe
stopping (p = 0.537), however the lower frequency of hypoglycemia (vs.
hyperglycemia) warrants a larger sample of drivers to investigate this effect.
Critically, presence of diabetes alone did not associate with unsafe stopping,
underscoring the need to evaluate driver physiology in licensing guidelines.
| stat |
Differential Privacy Dynamics of Langevin Diffusion and Noisy Gradient
Descent | We model the dynamics of privacy loss in Langevin diffusion and extend it to
the noisy gradient descent algorithm: we compute a tight bound on R\'enyi
differential privacy and the rate of its change throughout the learning
process. We prove that the privacy loss converges exponentially fast. This
significantly improves the prior privacy analysis of differentially private
(stochastic) gradient descent algorithms, where (R\'enyi) privacy loss
constantly increases over the training iterations. Unlike composition-based
methods in differential privacy, our privacy analysis does not assume that the
noisy gradients (or parameters) during the training could be revealed to the
adversary. Our analysis tracks the dynamics of privacy loss through the
algorithm's intermediate parameter distributions, thus allowing us to account
for privacy amplification due to convergence. We prove that our privacy
analysis is tight, and also provide a utility analysis for strongly convex,
smooth and Lipshitz loss functions.
| stat |
Modelling COVID-19 -- I A dynamic SIR(D) with application to Indian data | We propose an epidemiological model using an adaptive dynamic three
compartment (with four states) SIR(D) model. Our approach is similar to
non-parametric curve fitting in spirit and automatically adapts to key external
factors, such as interventions, while retaining the parsimonious nature of the
standard SIR(D) model. Initial dynamic temporal estimates of the model
parameters are obtained by minimising the aggregate residual sum of squares
across the number of infections, recoveries, and fatalities, over a chosen lag
period. Then a geometric smoother is applied to obtain the final time series of
estimates. These estimates are used to obtain dynamic temporal robust estimates
of the key feature of this pandemic, namely the "reproduction number". We
illustrate our method on the Indian COVID-19 data for the period March 14 -
August 31, 2020. The time series data plots of the 36 states and union
territories shows a clear presence of inter-regional variation in the prognosis
of the epidemic. This is also bourne out by the estimates of the underlying
parameters, including the reproduction numbers for the 36 regions. Due to this,
an SIR(D) model, dynamic or otherwise, on the national aggregate data is not
suited for robust local predictions. The time series of estimates of the model
enables us to carry out daily, weekly and also long term predictions, including
construction of predictive bands. We obtain an excellent agreement between the
actual data and the model predicted data at the regional level. Our estimates
of the current reproduction number turn out to be more than 2 in three regions
(Andhra Pradesh, Maharashtra and Uttar Pradesh) and between 1.5 and 2 in 13
regions. Each of these regions have experienced an individual trajectory, which
typically involves initial phase of shock(s) followed by a relatively steady
lower level of the reproduction number.
| stat |
Regression adjustments for estimating the global treatment effect in
experiments with interference | Standard estimators of the global average treatment effect can be biased in
the presence of interference. This paper proposes regression adjustment
estimators for removing bias due to interference in Bernoulli randomized
experiments. We use a fitted model to predict the counterfactual outcomes of
global control and global treatment. Our work differs from standard regression
adjustments in that the adjustment variables are constructed from functions of
the treatment assignment vector, and that we allow the researcher to use a
collection of any functions correlated with the response, turning the problem
of detecting interference into a feature engineering problem. We characterize
the distribution of the proposed estimator in a linear model setting and
connect the results to the standard theory of regression adjustments under
SUTVA. We then propose an estimator that allows for flexible machine learning
estimators to be used for fitting a nonlinear interference functional form. We
propose conducting statistical inference via bootstrap and resampling methods,
which allow us to sidestep the complicated dependences implied by interference
and instead rely on empirical covariance structures. Such variance estimation
relies on an exogeneity assumption akin to the standard unconfoundedness
assumption invoked in observational studies. In simulation experiments, our
methods are better at debiasing estimates than existing inverse propensity
weighted estimators based on neighborhood exposure modeling. We use our method
to reanalyze an experiment concerning weather insurance adoption conducted on a
collection of villages in rural China.
| stat |
Are You All Normal? It Depends! | The assumption of normality has underlain much of the development of
statistics, including spatial statistics, and many tests have been proposed. In
this work, we focus on the multivariate setting and we first provide a synopsis
of the recent advances in multivariate normality tests for i.i.d. data, with
emphasis on the skewness and kurtosis approaches. We show through simulation
studies that some of these tests cannot be used directly for testing normality
of spatial data, since the multivariate sample skewness and kurtosis measures,
such as the Mardia's measures, deviate from their theoretical values under
Gaussianity due to dependence, and some related tests exhibit inflated type I
error, especially when the spatial dependence gets stronger. We review briefly
the few existing tests under dependence (time or space), and then propose a new
multivariate normality test for spatial data by accounting for the spatial
dependence of the observations in the test statistic. The new test aggregates
univariate Jarque-Bera (JB) statistics, which combine skewness and kurtosis
measures, for individual variables. The asymptotic variances of sample skewness
and kurtosis for standardized observations are derived given the dependence
structure of the spatial data. Consistent estimators of the asymptotic
variances are then constructed for finite samples. The test statistic is easy
to compute, without any smoothing involved, and it is asymptotically
$\chi^2_{2p}$ under normality, where $p$ is the number of variables. The new
test has a good control of the type I error and a high empirical power,
especially for large sample sizes.
| stat |
Causal Impact of Web Browsing and Other Factors on Research Publications | In this paper, we study the causal impact of the web-search activity on the
research publication. We considered observational prospective study design,
where research activity of 267 scientists is being studied. We considered the
Poisson and negative binomial regression model for our analysis. Based on the
Akaike's Model selection criterion, we found the negative binomial regression
performs better than the Poisson regression. Detailed analysis indicates that
the higher web-search activity of 2016 related to the sci-indexed website has a
positive significant impact on the research publication of 2017. We observed
that unique collaborations of 2016 and web-search activity of 2016 have a
non-linear but significant positive impact on the research publication of 2017.
What-if analysis indicates the high web browsing activity leads to more number
of the publication. However, interestingly we see a scientist with low web
activity can be as productive as others if her/his maximum hits are the
sci-indexed journal. That is if the scientist uses web browsing only for
research-related activity, then she/he can be equally productive even if
her/his web activity is lower than fellow scientists.
| stat |
Graph representation forecasting of patient's medical conditions:
towards a digital twin | Objective: Modern medicine needs to shift from a wait and react, curative
discipline to a preventative, interdisciplinary science aiming at providing
personalised, systemic and precise treatment plans to patients. The aim of this
work is to present how the integration of machine learning approaches with
mechanistic computational modelling could yield a reliable infrastructure to
run probabilistic simulations where the entire organism is considered as a
whole. Methods: We propose a general framework that composes advanced AI
approaches and integrates mathematical modelling in order to provide a
panoramic view over current and future physiological conditions. The proposed
architecture is based on a graph neural network (GNNs) forecasting clinically
relevant endpoints (such as blood pressure) and a generative adversarial
network (GANs) providing a proof of concept of transcriptomic integrability.
Results: We show the results of the investigation of pathological effects of
overexpression of ACE2 across different signalling pathways in multiple tissues
on cardiovascular functions. We provide a proof of concept of integrating a
large set of composable clinical models using molecular data to drive local and
global clinical parameters and derive future trajectories representing the
evolution of the physiological state of the patient. Significance: We argue
that the graph representation of a computational patient has potential to solve
important technological challenges in integrating multiscale computational
modelling with AI. We believe that this work represents a step forward towards
a healthcare digital twin.
| stat |
Statistical Inference for Generative Models with Maximum Mean
Discrepancy | While likelihood-based inference and its variants provide a statistically
efficient and widely applicable approach to parametric inference, their
application to models involving intractable likelihoods poses challenges. In
this work, we study a class of minimum distance estimators for intractable
generative models, that is, statistical models for which the likelihood is
intractable, but simulation is cheap. The distance considered, maximum mean
discrepancy (MMD), is defined through the embedding of probability measures
into a reproducing kernel Hilbert space. We study the theoretical properties of
these estimators, showing that they are consistent, asymptotically normal and
robust to model misspecification. A main advantage of these estimators is the
flexibility offered by the choice of kernel, which can be used to trade-off
statistical efficiency and robustness. On the algorithmic side, we study the
geometry induced by MMD on the parameter space and use this to introduce a
novel natural gradient descent-like algorithm for efficient implementation of
these estimators. We illustrate the relevance of our theoretical results on
several classes of models including a discrete-time latent Markov process and
two multivariate stochastic differential equation models.
| stat |
Training a U-Net based on a random mode-coupling matrix model to recover
acoustic interference striations | A U-Net is trained to recover acoustic interference striations (AISs) from
distorted ones. A random mode-coupling matrix model is introduced to generate a
large number of training data quickly, which are used to train the U-Net. The
performance of AIS recovery of the U-Net is tested in range-dependent
waveguides with nonlinear internal waves (NLIWs). Although the random
mode-coupling matrix model is not an accurate physical model, the test results
show that the U-Net successfully recovers AISs under different signal-to-noise
ratios (SNRs) and different amplitudes and widths of NLIWs for different
shapes.
| stat |
Model-based Kernel Sum Rule: Kernel Bayesian Inference with
Probabilistic Models | Kernel Bayesian inference is a principled approach to nonparametric inference
in probabilistic graphical models, where probabilistic relationships between
variables are learned from data in a nonparametric manner. Various algorithms
of kernel Bayesian inference have been developed by combining kernelized basic
probabilistic operations such as the kernel sum rule and kernel Bayes' rule.
However, the current framework is fully nonparametric, and it does not allow a
user to flexibly combine nonparametric and model-based inferences. This is
inefficient when there are good probabilistic models (or simulation models)
available for some parts of a graphical model; this is in particular true in
scientific fields where "models" are the central topic of study. Our
contribution in this paper is to introduce a novel approach, termed the {\em
model-based kernel sum rule} (Mb-KSR), to combine a probabilistic model and
kernel Bayesian inference. By combining the Mb-KSR with the existing kernelized
probabilistic rules, one can develop various algorithms for hybrid (i.e.,
nonparametric and model-based) inferences. As an illustrative example, we
consider Bayesian filtering in a state space model, where typically there
exists an accurate probabilistic model for the state transition process. We
propose a novel filtering method that combines model-based inference for the
state transition process and data-driven, nonparametric inference for the
observation generating process. We empirically validate our approach with
synthetic and real-data experiments, the latter being the problem of
vision-based mobile robot localization in robotics, which illustrates the
effectiveness of the proposed hybrid approach.
| stat |
Wildfire Smoke and Air Quality: How Machine Learning Can Guide Forest
Management | Prescribed burns are currently the most effective method of reducing the risk
of widespread wildfires, but a largely missing component in forest management
is knowing which fuels one can safely burn to minimize exposure to toxic smoke.
Here we show how machine learning, such as spectral clustering and manifold
learning, can provide interpretable representations and powerful tools for
differentiating between smoke types, hence providing forest managers with vital
information on effective strategies to reduce climate-induced wildfires while
minimizing production of harmful smoke.
| stat |
Non-Linear Reduced Modeling by Generalized Kernel-Based Dynamic Mode
Decomposition | Reduced modeling of a computationally demanding dynamical system aims at
approximating its trajectories, while optimizing the trade-off between accuracy
and computational complexity. In this work, we propose to achieve such an
approximation by first embedding the trajectories in a reproducing kernel
Hilbert space (RKHS), which exhibits appealing approximation and computational
capabilities, and then solving the associated reduced model problem. More
specifically, we propose a new efficient algorithm for data-driven reduced
modeling of non-linear dynamics based on linear approximations in a RKHS. This
algorithm takes advantage of the closed-form solution of a low-rank constraint
optimization problem while exploiting advantageously kernel-based computations.
Reduced modeling with this algorithm reveals a gain in approximation accuracy,
as shown by numerical simulations, and in complexity with respect to existing
approaches.
| stat |
Bayesian approach to Lorenz curve using time series grouped data | This study is concerned with estimating the inequality measures associated
with the underlying hypothetical income distribution from the times series
grouped data on the Lorenz curve. We adopt the Dirichlet pseudo likelihood
approach where the parameters of the Dirichlet likelihood are set to the
differences between the Lorenz curve of the hypothetical income distribution
for the consecutive income classes and propose a state space model which
combines the transformed parameters of the Lorenz curve through a time series
structure. Furthermore, the information on the sample size in each survey is
introduced into the originally nuisance Dirichlet precision parameter to take
into account the variability from the sampling. From the simulated data and
real data on the Japanese monthly income survey, it is confirmed that the
proposed model produces more efficient estimates on the inequality measures
than the existing models without time series structures.
| stat |
Statistical Learning from Biased Training Samples | With the deluge of digitized information in the Big Data era, massive
datasets are becoming increasingly available for learning predictive models.
However, in many situations, the poor control of the data acquisition processes
may naturally jeopardize the outputs of machine-learning algorithms and
selection bias issues are now the subject of much attention in the literature.
It is precisely the purpose of the present article to investigate how to extend
Empirical Risk Minimization (ERM), the main paradigm of statistical learning,
when the training observations are generated from biased models, i.e. from
distributions that are different from that of the data in the test/prediction
stage. Precisely, we show how to build a "nearly debiased" training statistical
population from biased samples and the related biasing functions following in
the footsteps of the approach originally proposed in Vardi (1985). Furthermore,
we study from a non asymptotic perspective the performance of minimizers of an
empirical version of the risk computed from the statistical population thus
constructed. Remarkably, the learning rate achieved by this procedure is of the
same order as that attained in absence of any selection bias phenomenon. Beyond
these theoretical guarantees, illustrative experimental results supporting the
relevance of the algorithmic approach promoted in this paper are also
displayed.
| stat |
Constrained High Dimensional Statistical Inference | In typical high dimensional statistical inference problems, confidence
intervals and hypothesis tests are performed for a low dimensional subset of
model parameters under the assumption that the parameters of interest are
unconstrained. However, in many problems, there are natural constraints on
model parameters and one is interested in whether the parameters are on the
boundary of the constraint or not. e.g. non-negativity constraints for
transmission rates in network diffusion. In this paper, we provide algorithms
to solve this problem of hypothesis testing in high-dimensional statistical
models under constrained parameter space. We show that following our testing
procedure we are able to get asymptotic designed Type I error under the null.
Numerical experiments demonstrate that our algorithm has greater power than the
standard algorithms where the constraints are ignored. We demonstrate the
effectiveness of our algorithms on two real datasets where we have
{\emph{intrinsic}} constraint on the parameters.
| stat |
SCAU: Modeling spectral causality for multivariate time series with
applications to electroencephalograms | Electroencephalograms (EEG) are noninvasive measurement signals of electrical
neuronal activity in the brain. One of the current major statistical challenges
is formally measuring functional dependency between those complex signals. This
paper, proposes the spectral causality model (SCAU), a robust linear model,
under a causality paradigm, to reflect inter- and intra-frequency modulation
effects that cannot be identifiable using other methods. SCAU inference is
conducted with three main steps: (a) signal decomposition into frequency bins,
(b) intermediate spectral band mapping, and (c) dependency modeling through
frequency-specific autoregressive models (VAR). We apply SCAU to study complex
dependencies during visual and lexical fluency tasks (word generation and
visual fixation) in 26 participants' EEGs. We compared the connectivity
networks estimated using SCAU with respect to a VAR model. SCAU networks show a
clear contrast for both stimuli while the magnitude links also denoted a low
variance in comparison with the VAR networks. Furthermore, SCAU dependency
connections not only were consistent with findings in the neuroscience
literature, but it also provided further evidence on the directionality of the
spatio-spectral dependencies such as the delta-originated and theta-induced
links in the fronto-temporal brain network.
| stat |
MCMC-driven importance samplers | Monte Carlo methods are the standard procedure for estimating complicated
integrals of multidimensional Bayesian posterior distributions. In this work,
we focus on LAIS, a class of adaptive importance samplers where Markov chain
Monte Carlo (MCMC) algorithms are employed to drive an underlying multiple
importance sampling (IS) scheme. Its power lies in the simplicity of the
layered framework: the upper layer locates proposal densities by means of MCMC
algorithms; while the lower layer handles the multiple IS scheme, in order to
compute the final estimators. The modular nature of LAIS allows for different
possible choices in the upper and lower layers, that will have different
performance and computational costs. In this work, we propose different
enhancements in order to increase the efficiency and reduce the computational
cost, of both upper and lower layers. The different variants are essential if
we aim to address computational challenges arising in real-world applications,
such as highly concentrated posterior distributions (due to large amounts of
data, etc.). Hamiltonian-driven importance samplers are presented and tested.
Furthermore, we introduce different strategies for designing cheaper schemes,
for instance, recycling samples generated in the upper layer and using them in
the final estimators in the lower layer. Numerical experiments show the
benefits of the proposed schemes as compared to the vanilla version of LAIS and
other benchmark methods.
| stat |
Extending Mixture of Experts Model to Investigate Heterogeneity of
Trajectories: When, Where and How to Add Which Covariates | Researchers are usually interested in examining the impact of covariates when
separating heterogeneous samples into latent classes that are more homogeneous.
The majority of theoretical and empirical studies with such aims have focused
on identifying covariates as predictors of class membership in the structural
equation modeling framework. In other words, the covariates only indirectly
affect the sample heterogeneity. However, the covariates' influence on
between-individual differences can also be direct. This article presents a
mixture model that investigates covariates to explain within-cluster and
between-cluster heterogeneity simultaneously, known as a mixture-of-experts
(MoE) model. This study aims to extend the MoE framework to investigate
heterogeneity in nonlinear trajectories: to identify latent classes, covariates
as predictors to clusters, and covariates that explain within-cluster
differences in change patterns over time. Our simulation studies demonstrate
that the proposed model generally estimates the parameters unbiasedly,
precisely and exhibits appropriate empirical coverage for a nominal 95%
confidence interval. This study also proposes implementing structural equation
model forests to shrink the covariate space of the proposed mixture model. We
illustrate how to select covariates and construct the proposed model with
longitudinal mathematics achievement data. Additionally, we demonstrate that
the proposed mixture model can be further extended in the structural equation
modeling framework by allowing the covariates that have direct effects to be
time-varying.
| stat |
Learning a Compressed Sensing Measurement Matrix via Gradient Unrolling | Linear encoding of sparse vectors is widely popular, but is commonly
data-independent -- missing any possible extra (but a priori unknown) structure
beyond sparsity. In this paper we present a new method to learn linear encoders
that adapt to data, while still performing well with the widely used $\ell_1$
decoder. The convex $\ell_1$ decoder prevents gradient propagation as needed in
standard gradient-based training. Our method is based on the insight that
unrolling the convex decoder into $T$ projected subgradient steps can address
this issue. Our method can be seen as a data-driven way to learn a compressed
sensing measurement matrix. We compare the empirical performance of 10
algorithms over 6 sparse datasets (3 synthetic and 3 real). Our experiments
show that there is indeed additional structure beyond sparsity in the real
datasets; our method is able to discover it and exploit it to create excellent
reconstructions with fewer measurements (by a factor of 1.1-3x) compared to the
previous state-of-the-art methods. We illustrate an application of our method
in learning label embeddings for extreme multi-label classification, and
empirically show that our method is able to match or outperform the precision
scores of SLEEC, which is one of the state-of-the-art embedding-based
approaches.
| stat |
Sequential Learning of Principal Curves: Summarizing Data Streams on the
Fly | When confronted with massive data streams, summarizing data with dimension
reduction methods such as PCA raises theoretical and algorithmic pitfalls.
Principal curves act as a nonlinear generalization of PCA and the present paper
proposes a novel algorithm to automatically and sequentially learn principal
curves from data streams. We show that our procedure is supported by regret
bounds with optimal sublinear remainder terms. A greedy local search
implementation (called \texttt{slpc}, for Sequential Learning Principal Curves)
that incorporates both sleeping experts and multi-armed bandit ingredients is
presented, along with its regret computation and performance on synthetic and
real-life data.
| stat |
Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep
Generative Models | Learning generative models that span multiple data modalities, such as vision
and language, is often motivated by the desire to learn more useful,
generalisable representations that faithfully capture common underlying factors
between the modalities. In this work, we characterise successful learning of
such models as the fulfillment of four criteria: i) implicit latent
decomposition into shared and private subspaces, ii) coherent joint generation
over all modalities, iii) coherent cross-generation across individual
modalities, and iv) improved model learning for individual modalities through
multi-modal integration. Here, we propose a mixture-of-experts multimodal
variational autoencoder (MMVAE) to learn generative models on different sets of
modalities, including a challenging image-language dataset, and demonstrate its
ability to satisfy all four criteria, both qualitatively and quantitatively.
| stat |
Valid Instrumental Variables Selection Methods using Auxiliary Variable
and Constructing Efficient Estimator | In observational studies, we are usually interested in estimating causal
effects between treatments and outcomes. When some covariates are not observed,
an unbiased estimator usually cannot be obtained. In this paper, we focus on
instrumental variable (IV) methods. By using IVs, an unbiased estimator for
causal effects can be estimated even if there exists some unmeasured
covariates. Constructing a linear combination of IVs solves weak IV problems,
however, there are risks estimating biased causal effects by including some
invalid IVs. In this paper, we use Negative Control Outcomes as auxiliary
variables to select valid IVs. By using NCOs, there are no necessity to specify
not only the set of valid IVs but also invalid one in advance: this point is
different from previous methods. We prove that the estimated causal effects has
the same asymptotic variance as the estimator using Generalized Method of
Moments that has the semiparametric efficiency. Also, we confirm properties of
our method and previous methods through simulations.
| stat |
Do ideas have shape? Plato's theory of forms as the continuous limit of
artificial neural networks | We show that ResNets converge, in the infinite depth limit, to a
generalization of image registration algorithms. In this generalization, images
are replaced by abstractions (ideas) living in high dimensional RKHS spaces,
and material points are replaced by data points. Whereas computational anatomy
aligns images via deformations of the material space, this generalization
aligns ideas by via transformations of their RKHS. This identification of
ResNets as idea registration algorithms has several remarkable consequences.
The search for good architectures can be reduced to that of good kernels, and
we show that the composition of idea registration blocks with reduced
equivariant multi-channel kernels (introduced here) recovers and generalizes
CNNs to arbitrary spaces and groups of transformations. Minimizers of $L_2$
regularized ResNets satisfy a discrete least action principle implying the near
preservation of the norm of weights and biases across layers. The parameters of
trained ResNets can be identified as solutions of an autonomous Hamiltonian
system defined by the activation function and the architecture of the ANN.
Momenta variables provide a sparse representation of the parameters of a
ResNet. The registration regularization strategy provides a provably robust
alternative to Dropout for ANNs. Pointwise RKHS error estimates lead to
deterministic error estimates for ANNs.
| stat |
The Label Complexity of Active Learning from Observational Data | Counterfactual learning from observational data involves learning a
classifier on an entire population based on data that is observed conditioned
on a selection policy. This work considers this problem in an active setting,
where the learner additionally has access to unlabeled examples and can choose
to get a subset of these labeled by an oracle.
Prior work on this problem uses disagreement-based active learning, along
with an importance weighted loss estimator to account for counterfactuals,
which leads to a high label complexity. We show how to instead incorporate a
more efficient counterfactual risk minimizer into the active learning
algorithm. This requires us to modify both the counterfactual risk to make it
amenable to active learning, as well as the active learning process to make it
amenable to the risk. We provably demonstrate that the result of this is an
algorithm which is statistically consistent as well as more label-efficient
than prior work.
| stat |
The Impact of Extraneous Variables on the Performance of Recurrent
Neural Network Models in Clinical Tasks | Electronic Medical Records (EMR) are a rich source of patient information,
including measurements reflecting physiologic signs and administered therapies.
Identifying which variables are useful in predicting clinical outcomes can be
challenging. Advanced algorithms such as deep neural networks were designed to
process high-dimensional inputs containing variables in their measured form,
thus bypass separate feature selection or engineering steps. We investigated
the effect of extraneous input variables on the predictive performance of
Recurrent Neural Networks (RNN) by including in the input vector extraneous
variables randomly drawn from theoretical and empirical distributions. RNN
models using different input vectors (EMR variables; EMR and extraneous
variables; extraneous variables only) were trained to predict three clinical
outcomes: in-ICU mortality, 72-hour ICU re-admission, and 30-day ICU-free days.
The measured degradations of the RNN's predictive performance with the addition
of extraneous variables to EMR variables were negligible.
| stat |
Asymmetric Tobit analysis for correlation estimation from censored data | Contamination of water resources with pathogenic microorganisms excreted in
human feces is a worldwide public health concern. Surveillance of fecal
contamination is commonly performed by routine monitoring for a single type or
a few types of microorganism(s). To design a feasible routine for periodic
monitoring and to control risks of exposure to pathogens, reliable statistical
algorithms for inferring correlations between concentrations of microorganisms
in water need to be established. Moreover, because pathogens are often present
in low concentrations, some contaminations are likely to be under a detection
limit. This yields a pairwise left-censored dataset and complicates computation
of correlation coefficients. Errors of correlation estimation can be smaller if
undetected values are imputed better. To obtain better imputations, we utilize
side information and develop a new technique, the \emph{asymmetric Tobit model}
which is an extension of the Tobit model so that domain knowledge can be
exploited effectively when fitting the model to a censored dataset. The
empirical results demonstrate that imputation with domain knowledge is
effective for this task.
| stat |
A Variational Expectation-Maximisation Algorithm for Learning Jump
Markov Linear Systems | Jump Markov linear systems (JMLS) are a useful class which can be used to
model processes which exhibit random changes in behavior during operation. This
paper presents a numerically stable method for learning the parameters of jump
Markov linear systems using the expectation-maximisation (EM) approach. The
solution provided herein is a deterministic algorithm, and is not a Monte Carlo
based technique. As a result, simulations show that when compared to
alternative approaches, a more likely set of system parameters can be found
within a fixed computation time, which better explain the observations of the
system.
| stat |
Lessons from the German Tank Problem | During World War II the German army used tanks to devastating advantage. The
Allies needed accurate estimates of their tank production and deployment. They
used two approaches to find these values: spies, and statistics. This note
describes the statistical approach. Assuming the tanks are labeled
consecutively starting at 1, if we observe $k$ serial numbers from an unknown
number $N$ of tanks, with the maximum observed value $m$, then the best
estimate for $N$ is $m(1 + 1/k) - 1$. This is now known as the German Tank
Problem, and is a terrific example of the applicability of mathematics and
statistics in the real world. The first part of the paper reproduces known
results, specifically deriving this estimate and comparing its effectiveness to
that of the spies. The second part presents a result we have not found in print
elsewhere, the generalization to the case where the smallest value is not
necessarily 1. We emphasize in detail why we are able to obtain such clean,
closed-form expressions for the estimates, and conclude with an appendix
highlighting how to use this problem to teach regression and how statistics can
help us find functional relationships.
| stat |
Efficient Data Representation by Selecting Prototypes with Importance
Weights | Prototypical examples that best summarizes and compactly represents an
underlying complex data distribution communicate meaningful insights to humans
in domains where simple explanations are hard to extract. In this paper we
present algorithms with strong theoretical guarantees to mine these data sets
and select prototypes a.k.a. representatives that optimally describes them. Our
work notably generalizes the recent work by Kim et al. (2016) where in addition
to selecting prototypes, we also associate non-negative weights which are
indicative of their importance. This extension provides a single coherent
framework under which both prototypes and criticisms (i.e. outliers) can be
found. Furthermore, our framework works for any symmetric positive definite
kernel thus addressing one of the key open questions laid out in Kim et al.
(2016). By establishing that our objective function enjoys a key property of
that of weak submodularity, we present a fast ProtoDash algorithm and also
derive approximation guarantees for the same. We demonstrate the efficacy of
our method on diverse domains such as retail, digit recognition (MNIST) and on
publicly available 40 health questionnaires obtained from the Center for
Disease Control (CDC) website maintained by the US Dept. of Health. We validate
the results quantitatively as well as qualitatively based on expert feedback
and recently published scientific studies on public health, thus showcasing the
power of our technique in providing actionability (for retail), utility (for
MNIST) and insight (on CDC datasets) which arguably are the hallmarks of an
effective data mining method.
| stat |
Accelerating System Adequacy Assessment using the Multilevel Monte Carlo
Approach | Accurately and efficiently estimating system performance under uncertainty is
paramount in power system planning and operation. Monte Carlo simulation is
often used for this purpose, but convergence may be slow, especially when
detailed models are used. Previously published methods to speed up computations
may severely constrain model complexity, limiting their real-world
effectiveness. This paper uses the recently proposed Multilevel Monte Carlo
(MLMC) framework, which combines outputs from a hierarchy of simulators to
boost computational efficiency without sacrificing accuracy. It explains which
requirements the MLMC framework imposes on the model hierarchy, and how these
naturally occur in power system adequacy assessment problems. Two adequacy
assessment examples are studied in detail: a composite system and a system with
heterogeneous storage units. An intuitive speed metric is introduced for easy
comparison of simulation setups. Depending on the problem and metric of
interest, large speedups can be obtained.
| stat |
Combining Broad and Narrow Case Definitions in Matched Case-Control
Studies | In a matched case-control study, cases are compared to noncases, who are
similar in observed covariates, in terms of their retrospective exposure to a
treatment to assess the impact of the treatment on the outcome. In the absence
of a gold standard case definition, there is often a broad case definition and
a narrow case definition. The broad case definition offers a larger sample size
of cases but the narrow case definition may offer a larger effect size.
Restricting to the narrow case definition may introduce selection bias because
the treatment may affect the type of a case a subject is. In this article, we
propose a new sensitivity analysis framework for combining broad and narrow
case definitions in matched case-control studies, that considers the unmeasured
confounding bias and selection bias simultaneously. We develop a valid
randomization-based testing procedure using only the narrow case matched sets
when the effect of the unmeasured confounder on receiving treatment and the
effect of the treatment on case definition among the always-cases are
controlled by sensitivity parameters. We then use the Bonferroni method to
combine the testing procedures using the broad and narrow case definitions. We
also study comprehensively the proposed testing procedures' sensitivity to
unmeasured biases using the design sensitivity and extensive power analyses.
Our method is applied to study whether having firearms at home increases
suicide risk.
| stat |
Interval estimators for inequality measures using grouped data | Income inequality measures are often used as an indication of economic
health. How to obtain reliable confidence intervals for these measures based on
sampled data has been studied extensively in recent years. To preserve
confidentiality, income data is often made available in summary form only (i.e.
histograms, frequencies between quintiles, etc.). In this paper, we show that
good coverage can be achieved for bootstrap and Wald-type intervals for
quantile-based measures when only grouped (binned) data are available. These
coverages are typically superior to those that we have been able to achieve for
intervals for popular measures such as the Gini index in this grouped data
setting. To facilitate the bootstrapping, we use the Generalized Lambda
Distribution and also a linear interpolation approximation method to
approximate the underlying density. The latter is possible when groups means
are available. We also apply our methods to real data sets.
| stat |
$\P$ILCRO: Making Importance Landscapes Flat Again | Convolutional neural networks have had a great success in numerous tasks,
including image classification, object detection, sequence modelling, and many
more. It is generally assumed that such neural networks are translation
invariant, meaning that they can detect a given feature independent of its
location in the input image. While this is true for simple cases, where
networks are composed of a restricted number of layer classes and where images
are fairly simple, complex images with common state-of-the-art networks do not
usually enjoy this property as one might hope. This paper shows that most of
the existing convolutional architectures define, at initialisation, a specific
feature importance landscape that conditions their capacity to attend to
different locations of the images later during training or even at test time.
We demonstrate how this phenomenon occurs under specific conditions and how it
can be adjusted under some assumptions. We derive the P-objective, or PILCRO
for Pixel-wise Importance Landscape Curvature Regularised Objective, a simple
regularisation technique that favours weight configurations that produce
smooth, low-curvature importance landscapes that are conditioned on the data
and not on the chosen architecture. Through extensive experiments, we further
show that P-regularised versions of popular computer vision networks have a
flat importance landscape, train faster, result in a better accuracy and are
more robust to noise at test time, when compared to their original counterparts
in common computer-vision classification settings.
| stat |
A modified risk detection approach of biomarkers by frailty effect on
multiple time to event data | Multiple indications of disease progression found in a cancer patient by
loco-regional relapse, distant metastasis and death. Early identification of
these indications is necessary to change the treatment strategy. Biomarkers
play an essential role in this aspect. The biomarkers can influence how
particular cancer behaves and how it may respond to a specific treatment. The
survival chance of a patient is dependent on the biomarker, and the treatment
strategy also differs accordingly, e.g., the survival prediction of breast
cancer patients diagnosed with HER2 positive status is different from the same
with HER2 negative status. This results in a different treatment strategy. So,
the heterogeneity of the biomarker statuses or levels should be taken into
consideration while modelling the survival outcome. This heterogeneity factor
which is often unobserved, is called frailty. When multiple indications are
present simultaneously, the scenario becomes more complex as only one of them
can occur, which will censor the occurrence of other events. The events
indicating cancer progression are likely to be inter-related. So, the
correlation should be incorporated through the frailties of different events.
In our study, we considered a multiple events or risks model with a
heterogeneity component. Based on the estimated variance of the frailty, the
threshold levels of a biomarker are utilised as early detection tool of the
disease progression or death. Additive-gamma frailty model is considered to
account the correlation between different frailty components and estimation of
parameters are performed using Expectation-Maximization Algorithm. With the
extensive algorithm in R, we have obtained the threshold levels of activity of
a biomarker in a multiple events scenario.
| stat |
Optimal tests for elliptical symmetry: specified and unspecified
location | Although the assumption of elliptical symmetry is quite common in
multivariate analysis and widespread in a number of applications, the problem
of testing the null hypothesis of ellipticity so far has not been addressed in
a fully satisfactory way. Most of the literature in the area indeed addresses
the null hypothesis of elliptical symmetry with specified location and actually
addresses location rather than non-elliptical alternatives. In this paper, we
are proposing new classes of testing procedures, both for specified and
unspecified location. The backbone of our construction is Le Cam's asymptotic
theory of statistical experiments, and optimality is to be understood locally
and asymptotically within the family of generalized skew-elliptical
distributions. The tests we are proposing are meeting all the desired
properties of a``good'' test of elliptical symmetry: they have a simple
asymptotic distribution under the entire null hypothesis of elliptical symmetry
with unspecified radial density and shape parameter; they are affine-invariant,
computationally fast, intuitively understandable, and not too demanding in
terms of moments. While achieving optimality against generalized
skew-elliptical alternatives, they remain quite powerful under a much broader
class of non-elliptical distributions and significantly outperform the
available competitors.
| stat |
New estimation approaches for graphical models with elastic net penalty | In the context of undirected Gaussian graphical models, we introduce three
estimators based on elastic net penalty for the underlying dependence graph.
Our goal is to estimate the sparse precision matrix, from which to retrieve
both the underlying conditional dependence graph and the partial correlation
graph. The first estimator is derived from the direct penalization of the
precision matrix in the likelihood function, while the second from using
conditional penalized regressions to estimate the precision matrix. Finally,
the third estimator relies on a 2-stages procedure that estimates the edge set
first and then the precision matrix elements. Through simulations we
investigate the performances of the proposed methods on a large set of
well-known network structures. Empirical results on simulated data show that
the 2-stages procedure outperforms all other estimators both w.r.t. estimating
the sparsity pattern in the graph and the edges' weights. Finally, using
real-world data on US economic sectors, we estimate dependencies and show the
impact of Covid-19 pandemic on the network strength.
| stat |
Penalized Poisson model for network meta-analysis of individual patient
time-to-event data | Network meta-analysis (NMA) allows the combination of direct and indirect
evidence from a set of randomized clinical trials. Performing NMA using
individual patient data (IPD) is considered as a gold standard approach as it
provides several advantages over NMA based on aggregate data. For example, it
allows to perform advanced modelling of covariates or covariate-treatment
interactions. An important issue in IPD NMA is the selection of influential
parameters among terms that account for inconsistency, covariates,
covariate-by-treatment interactions or non-proportionality of treatments effect
for time to event data. This issue has not been deeply studied in the
literature yet and in particular not for time-to-event data. A major difficulty
is to jointly account for between-trial heterogeneity which could have a major
influence on the selection process. The use of penalized generalized mixed
effect model is a solution, but existing implementations have several
shortcomings and an important computational cost that precludes their use for
complex IPD NMA. In this article, we propose a penalized Poisson regression
model to perform IPD NMA of time-to-event data. It is based only on fixed
effect parameters which improve its computational cost over the use of random
effects. It could be easily implemented using existing penalized regression
package. Computer code is shared for implementation. The methods were applied
on simulated data to illustrate the importance to take into account between
trial heterogeneity during the selection procedure. Finally, it was applied to
an IPD NMA of overall survival of chemotherapy and radiotherapy in
nasopharyngeal carcinoma.
| stat |
Varying impacts of letters of recommendation on college admissions:
Approximate balancing weights for subgroup effects in observational studies | In a pilot program during the 2016-17 admissions cycle, the University of
California, Berkeley invited many applicants for freshman admission to submit
letters of recommendation. We use this pilot as the basis for an observational
study of the impact of submitting letters of recommendation on subsequent
admission, with the goal of estimating how impacts vary across pre-defined
subgroups. Understanding this variation is challenging in observational
studies, however, because estimated impacts reflect both actual treatment
effect variation and differences in covariate balance across groups. To address
this, we develop balancing weights that directly optimize for ``local balance''
within subgroups while maintaining global covariate balance between treated and
control units. We then show that this approach has a dual representation as a
form of inverse propensity score weighting with a hierarchical propensity score
model. In the UC Berkeley pilot study, our proposed approach yields excellent
local and global balance, unlike more traditional weighting methods, which fail
to balance covariates within subgroups. We find that the impact of letters of
recommendation increases with the predicted probability of admission, with
mixed evidence of differences for under-represented minority applicants.
| stat |
Estimating Perinatal Critical Windows to Environmental Mixtures via
Structured Bayesian Regression Tree Pairs | Maternal exposure to environmental chemicals during pregnancy can alter birth
and children's health outcomes. Research seeks to identify critical windows,
time periods when the exposures can change future health outcomes, and estimate
the exposure-response relationship. Existing statistical approaches focus on
estimation of the association between maternal exposure to a single
environmental chemical observed at high-temporal resolution, such as weekly
throughout pregnancy, and children's health outcomes. Extending to multiple
chemicals observed at high temporal resolution poses a dimensionality problem
and statistical methods are lacking. We propose a tree-based model for mixtures
of exposures that are observed at high temporal resolution. The proposed
approach uses an additive ensemble of structured tree-pairs that define
structured main effects and interactions between time-resolved predictors and
variable selection to select out of the model predictors not correlated with
the outcome. We apply our method in a simulation and the analysis of the
relationship between five exposures measured weekly throughout pregnancy and
resulting birth weight in a Denver, Colorado birth cohort. We identified
critical windows during which fine particulate matter, sulfur dioxide, and
temperature are negatively associated with birth weight and an interaction
between fine particulate matter and temperature. Software is made available in
the R package dlmtree.
| stat |
Generative Ratio Matching Networks | Deep generative models can learn to generate realistic-looking images, but
many of the most effective methods are adversarial and involve a saddlepoint
optimization, which requires a careful balancing of training between a
generator network and a critic network. Maximum mean discrepancy networks
(MMD-nets) avoid this issue by using kernel as a fixed adversary, but
unfortunately, they have not on their own been able to match the generative
quality of adversarial training. In this work, we take their insight of using
kernels as fixed adversaries further and present a novel method for training
deep generative models that does not involve saddlepoint optimization. We call
our method generative ratio matching or GRAM for short. In GRAM, the generator
and the critic networks do not play a zero-sum game against each other,
instead, they do so against a fixed kernel. Thus GRAM networks are not only
stable to train like MMD-nets but they also match and beat the generative
quality of adversarially trained generative networks.
| stat |
Sampling strategies for fast updating of Gaussian Markov random fields | Gaussian Markov random fields (GMRFs) are popular for modeling dependence in
large areal datasets due to their ease of interpretation and computational
convenience afforded by the sparse precision matrices needed for random
variable generation. Typically in Bayesian computation, GMRFs are updated
jointly in a block Gibbs sampler or componentwise in a single-site sampler via
the full conditional distributions. The former approach can speed convergence
by updating correlated variables all at once, while the latter avoids solving
large matrices. We consider a sampling approach in which the underlying graph
can be cut so that conditionally independent sites are updated simultaneously.
This algorithm allows a practitioner to parallelize updates of subsets of
locations or to take advantage of `vectorized' calculations in a high-level
language such as R. Through both simulated and real data, we demonstrate
computational savings that can be achieved versus both single-site and block
updating, regardless of whether the data are on a regular or an irregular
lattice. The approach provides a good compromise between statistical and
computational efficiency and is accessible to statisticians without expertise
in numerical analysis or advanced computing.
| stat |
Rapid Detection of Hot-spot by Tensor Decomposition with Application to
Weekly Gonorrhea Data | In many bio-surveillance and healthcare applications, data sources are
measured from many spatial locations repeatedly over time, say,
daily/weekly/monthly. In these applications, we are typically interested in
detecting hot-spots, which are defined as some structured outliers that are
sparse over the spatial domain but persistent over time. In this paper, we
propose a tensor decomposition method to detect when and where the hot-spots
occur. Our proposed methods represent the observed raw data as a
three-dimensional tensor including a circular time dimension for
daily/weekly/monthly patterns, and then decompose the tensor into three
components: smooth global trend, local hot-spots, and residuals. A combination
of LASSO and fused LASSO is used to estimate the model parameters, and a CUSUM
procedure is applied to detect when and where the hot-spots might occur. The
usefulness of our proposed methodology is validated through numerical
simulation and a real-world dataset in the weekly number of gonorrhea cases
from $2006$ to $2018$ for $50$ states in the United States.
| stat |
Kernel Thinning | We introduce kernel thinning, a simple algorithm for generating
better-than-Monte-Carlo approximations to distributions $\mathbb{P}$ on
$\mathbb{R}^d$. Given $n$ input points, a suitable reproducing kernel
$\mathbf{k}$, and $\mathcal{O}(n^2)$ time, kernel thinning returns $\sqrt{n}$
points with comparable integration error for every function in the associated
reproducing kernel Hilbert space. With high probability, the maximum
discrepancy in integration error is $\mathcal{O}_d(n^{-\frac{1}{2}}\sqrt{\log
n})$ for compactly supported $\mathbb{P}$ and $\mathcal{O}_d(n^{-\frac{1}{2}}
\sqrt{(\log n)^{d+1}\log\log n})$ for sub-exponential $\mathbb{P}$. In
contrast, an equal-sized i.i.d. sample from $\mathbb{P}$ suffers
$\Omega(n^{-\frac14})$ integration error. Our sub-exponential guarantees
resemble the classical quasi-Monte Carlo error rates for uniform $\mathbb{P}$
on $[0,1]^d$ but apply to general distributions on $\mathbb{R}^d$ and a wide
range of common kernels. We use our results to derive explicit non-asymptotic
maximum mean discrepancy bounds for Gaussian, Mat\'ern, and B-spline kernels
and present two vignettes illustrating the practical benefits of kernel
thinning over i.i.d. sampling and standard Markov chain Monte Carlo thinning.
| stat |
Bayesian posterior approximation via greedy particle optimization | In Bayesian inference, the posterior distributions are difficult to obtain
analytically for complex models such as neural networks. Variational inference
usually uses a parametric distribution for approximation, from which we can
easily draw samples. Recently discrete approximation by particles has attracted
attention because of its high expression ability. An example is Stein
variational gradient descent (SVGD), which iteratively optimizes particles.
Although SVGD has been shown to be computationally efficient empirically, its
theoretical properties have not been clarified yet and no finite sample bound
of the convergence rate is known. Another example is the Stein points (SP)
method, which minimizes kernelized Stein discrepancy directly. Although a
finite sample bound is assured theoretically, SP is computationally inefficient
empirically, especially in high-dimensional problems. In this paper, we propose
a novel method named maximum mean discrepancy minimization by the Frank-Wolfe
algorithm (MMD-FW), which minimizes MMD in a greedy way by the FW algorithm.
Our method is computationally efficient empirically and we show that its finite
sample convergence bound is in a linear order in finite dimensions.
| stat |
Entropy Partial Transport with Tree Metrics: Theory and Practice | Optimal transport (OT) theory provides powerful tools to compare probability
measures. However, OT is limited to nonnegative measures having the same mass,
and suffers serious drawbacks about its computation and statistics. This leads
to several proposals of regularized variants of OT in the recent literature. In
this work, we consider an \textit{entropy partial transport} (EPT) problem for
nonnegative measures on a tree having different masses. The EPT is shown to be
equivalent to a standard complete OT problem on a one-node extended tree. We
derive its dual formulation, then leverage this to propose a novel
regularization for EPT which admits fast computation and negative definiteness.
To our knowledge, the proposed regularized EPT is the first approach that
yields a \textit{closed-form} solution among available variants of unbalanced
OT. For practical applications without priori knowledge about the tree
structure for measures, we propose tree-sliced variants of the regularized EPT,
computed by averaging the regularized EPT between these measures using random
tree metrics, built adaptively from support data points. Exploiting the
negative definiteness of our regularized EPT, we introduce a positive definite
kernel, and evaluate it against other baselines on benchmark tasks such as
document classification with word embedding and topological data analysis. In
addition, we empirically demonstrate that our regularization also provides
effective approximations.
| stat |
Evaluating Fairness Using Permutation Tests | Machine learning models are central to people's lives and impact society in
ways as fundamental as determining how people access information. The gravity
of these models imparts a responsibility to model developers to ensure that
they are treating users in a fair and equitable manner. Before deploying a
model into production, it is crucial to examine the extent to which its
predictions demonstrate biases. This paper deals with the detection of bias
exhibited by a machine learning model through statistical hypothesis testing.
We propose a permutation testing methodology that performs a hypothesis test
that a model is fair across two groups with respect to any given metric. There
are increasingly many notions of fairness that can speak to different aspects
of model fairness. Our aim is to provide a flexible framework that empowers
practitioners to identify significant biases in any metric they wish to study.
We provide a formal testing mechanism as well as extensive experiments to show
how this method works in practice.
| stat |
Confidence intervals with maximal average power | We propose a frequentist testing procedure that maintains a defined coverage
and is optimal in the sense that it gives maximal power to detect deviations
from a null hypothesis when the alternative to the null hypothesis is sampled
from a pre-specified distribution (the prior distribution). Selecting a prior
distribution allows to tune the decision rule. This leads to an increased
power, if the true data generating distribution happens to be compatible with
the prior. It comes at the cost of losing power, if the data generating
distribution or the observed data are incompatible with the prior. We
illustrate the proposed approach for a binomial experiment, which is
sufficiently simple such that the decision sets can be illustrated in figures,
which should facilitate an intuitive understanding. The potential beyond the
simple example will be discussed: the approach is generic in that the test is
defined based on the likelihood function and the prior only. It is
comparatively simple to implement and efficient to execute, since it does not
rely on Minimax optimization. Conceptually it is interesting to note that for
constructing the testing procedure the Bayesian posterior probability
distribution is used.
| stat |
A sensitivity analysis of the PAWN sensitivity index | The PAWN index is gaining traction among the modelling community as a
sensitivity measure. However, the robustness to its design parameters has not
yet been scrutinized: the size ($N$) and sampling ($\varepsilon$) of the model
output, the number of conditioning intervals ($n$) or the summary statistic
($\theta$). Here we fill this gap by running a sensitivity analysis of a
PAWN-based sensitivity analysis. We compare the results with the design
uncertainties of the Sobol' total-order index ($S_{Ti}^*$). Unlike in
$S_{Ti}^*$, the design uncertainties in PAWN create non-negligible chances of
producing biased results when ranking or screening inputs. The dependence of
PAWN upon ($N,n,\varepsilon, \theta$) is difficult to tame, as these parameters
interact with one another. Even in an ideal setting in which the optimum choice
for ($N,n,\varepsilon, \theta$) is known in advance, PAWN might not allow to
distinguish an influential, non-additive model input from a truly
non-influential model input.
| stat |
A Comparison of Methods for Identifying Location Effects in Unreplicated
Fractional Factorials in the Presence of Dispersion Effects | Most methods for identifying location effects in unreplicated fractional
factorial designs assume homoscedasticity of the response values. However,
dispersion effects in the underlying process may create heteroscedasticity in
the response values. This heteroscedasticity may go undetected when
identification of location effects is pursued. Indeed, methods for identifying
dispersion effects typically require first modeling location effects.
Therefore, it is imperative to understand how methods for identifying location
effects function in the presence of undetected dispersion effects. We used
simulation studies to examine the robustness of four different methods for
identifying location effects---Box and Meyer (1986), Lenth (1989), Berk and
Picard (1991), and Loughin and Noble (1997)---under models with one, two, or
three dispersion effects of varying sizes. We found that the first three
methods usually performed acceptably with respect to error rates and power, but
the Loughin-Noble method lost control of the individual error rate when
moderate-to-large dispersion effects were present.
| stat |
A Unified Framework for Structured Graph Learning via Spectral
Constraints | Graph learning from data represents a canonical problem that has received
substantial attention in the literature. However, insufficient work has been
done in incorporating prior structural knowledge onto the learning of
underlying graphical models from data. Learning a graph with a specific
structure is essential for interpretability and identification of the
relationships among data. Useful structured graphs include the multi-component
graph, bipartite graph, connected graph, sparse graph, and regular graph. In
general, structured graph learning is an NP-hard combinatorial problem,
therefore, designing a general tractable optimization method is extremely
challenging. In this paper, we introduce a unified graph learning framework
lying at the integration of Gaussian graphical models and spectral graph
theory. To impose a particular structure on a graph, we first show how to
formulate the combinatorial constraints as an analytical property of the graph
matrix. Then we develop an optimization framework that leverages graph learning
with specific structures via spectral constraints on graph matrices. The
proposed algorithms are provably convergent, computationally efficient, and
practically amenable for numerous graph-based tasks. Extensive numerical
experiments with both synthetic and real data sets illustrate the effectiveness
of the proposed algorithms. The code for all the simulations is made available
as an open source repository.
| stat |
Improving Molecular Graph Neural Network Explainability with
Orthonormalization and Induced Sparsity | Rationalizing which parts of a molecule drive the predictions of a molecular
graph convolutional neural network (GCNN) can be difficult. To help, we propose
two simple regularization techniques to apply during the training of GCNNs:
Batch Representation Orthonormalization (BRO) and Gini regularization. BRO,
inspired by molecular orbital theory, encourages graph convolution operations
to generate orthonormal node embeddings. Gini regularization is applied to the
weights of the output layer and constrains the number of dimensions the model
can use to make predictions. We show that Gini and BRO regularization can
improve the accuracy of state-of-the-art GCNN attribution methods on artificial
benchmark datasets. In a real-world setting, we demonstrate that medicinal
chemists significantly prefer explanations extracted from regularized models.
While we only study these regularizers in the context of GCNNs, both can be
applied to other types of neural networks
| stat |
A Hierarchical Bayes Unit-Level Small Area Estimation Model for Normal
Mixture Populations | National statistical agencies are regularly required to produce estimates
about various subpopulations, formed by demographic and/or geographic
classifications, based on a limited number of samples. Traditional direct
estimates computed using only sampled data from individual subpopulations are
usually unreliable due to small sample sizes. Subpopulations with small samples
are termed small areas or small domains. To improve on the less reliable direct
estimates, model-based estimates, which borrow information from suitable
auxiliary variables, have been extensively proposed in the literature. However,
standard model-based estimates rely on the normality assumptions of the error
terms. In this research we propose a hierarchical Bayesian (HB) method for the
unit-level nested error regression model based on a normal mixture for the
unit-level error distribution. To implement our proposal we use a uniform prior
for the regression parameters, random effects variance parameter, and the
mixing proportion, and we use a partially proper non-informative prior
distribution for the two unit-level error variance components in the mixture.
We apply our method to two examples to predict summary characteristics of farm
products at the small area level. One of the examples is prediction of twelve
county-level crop areas cultivated for corn in some Iowa counties. The other
example involves total cash associated in farm operations in twenty-seven
farming regions in Australia. We compare predictions of small area
characteristics based on the proposed method with those obtained by applying
the Datta and Ghosh (1991) and the Chakraborty et al. (2018) HB methods. Our
simulation study comparing these three Bayesian methods showed the superiority
of our proposed method, measured by prediction mean squared error, coverage
probabilities and lengths of credible intervals for the small area means.
| stat |
An algorithm to compute the $t$-value of a digital net and of its
projections | Digital nets are among the most successful methods to construct
low-discrepancy point sets for quasi-Monte Carlo integration. Their quality is
traditionally assessed by a measure called the $t$-value. A refinement computes
the $t$-value of the projections over subsets of coordinates and takes a
weighted average (or some other function) of these values. It is also of
interest to compute the $t$-values of embedded nets obtained by taking subsets
of the points. In this paper, we propose an efficient algorithm to compute such
measures and we compare our approach with previously proposed methods both
empirically and in terms of computational complexity.
| stat |
Sparse tree-based clustering of microbiome data to characterize
microbiome heterogeneity in pancreatic cancer | There is a keen interest in characterizing variation in the microbiome across
cancer patients, given increasing evidence of its important role in determining
treatment outcomes. Here our goal is to discover subgroups of patients with
similar microbiome profiles. We propose a novel unsupervised clustering
approach in the Bayesian framework that innovates over existing model-based
clustering approaches, such as the Dirichlet multinomial mixture model, in
three key respects: we incorporate feature selection, learn the appropriate
number of clusters from the data, and integrate information on the tree
structure relating the observed features. We compare the performance of our
proposed method to existing methods on simulated data designed to mimic real
microbiome data. We then illustrate results obtained for our motivating data
set, a clinical study aimed at characterizing the tumor microbiome of
pancreatic cancer patients.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.