title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Semiparametric inference on general functionals of two semicontinuous
populations | In this paper, we propose new semiparametric procedures for making inference
on linear functionals and their functions of two semicontinuous populations.
The distribution of each population is usually characterized by a mixture of a
discrete point mass at zero and a continuous skewed positive component, and
hence such distribution is semicontinuous in the nature. To utilize the
information from both populations, we model the positive components of the two
mixture distributions via a semiparametric density ratio model. Under this
model setup, we construct the maximum empirical likelihood estimators of the
linear functionals and their functions, and establish the asymptotic normality
of the proposed estimators. We show the proposed estimators of the linear
functionals are more efficient than the fully nonparametric ones. The developed
asymptotic results enable us to construct confidence regions and perform
hypothesis tests for the linear functionals and their functions. We further
apply these results to several important summary quantities such as the
moments, the mean ratio, the coefficient of variation, and the generalized
entropy class of inequality measures. Simulation studies demonstrate the
advantages of our proposed semiparametric method over some existing methods.
Two real data examples are provided for illustration.
| stat |
An information criterion for auxiliary variable selection in incomplete
data analysis | Statistical inference is considered for variables of interest, called primary
variables, when auxiliary variables are observed along with the primary
variables. We consider the setting of incomplete data analysis, where some
primary variables are not observed. Utilizing a parametric model of joint
distribution of primary and auxiliary variables, it is possible to improve the
estimation of parametric model for the primary variables when the auxiliary
variables are closely related to the primary variables. However, the estimation
accuracy reduces when the auxiliary variables are irrelevant to the primary
variables. For selecting useful auxiliary variables, we formulate the problem
as model selection, and propose an information criterion for predicting primary
variables by leveraging auxiliary variables. The proposed information criterion
is an asymptotically unbiased estimator of the Kullback-Leibler divergence for
complete data of primary variables under some reasonable conditions. We also
clarify an asymptotic equivalence between the proposed information criterion
and a variant of leave-one-out cross validation. Performance of our method is
demonstrated via a simulation study and a real data example.
| stat |
Visualizing and Understanding Large-Scale Assessments in Mathematics
through Dimensionality Reduction | In this paper, we apply the Logistic PCA (LPCA) as a dimensionality reduction
tool for visualizing patterns and characterizing the relevance of mathematics
abilities from a given population measured by a large-scale assessment. We
establish an equivalence of parameters between LPCA, Inner Product
Representation (IPR) and the two paramenter logistic model (2PL) from the Item
Response Theory (IRT). This equivalence provides three complemetary ways of
looking at data that assists professionals in education to perform in-context
interpretations. Particularly, we analyse the data collected from SPAECE, a
large-scale assessment in Mathematics that has been applied yearly in the
public educational system of the state of Cear\'a, Brazil. As the main result,
we show that the the poor performance of examinees in the end of middle school
is primarily caused by their disabilities in number sense.
| stat |
A unified model of inspection and monitoring quality | Non-destructive evaluation (NDE) through inspection and monitoring is an
integral part of asset integrity management. The relationship between the
condition of interest and the quantity measured by NDE is described with
probabilistic models such as PoD or ROC curves. These models are used to assess
the quality of the information provided by NDE systems, which is affected by
factors such as the experience of the inspector, environmental conditions, ease
of access, or imprecision in the measuring device. In this paper, we show how
the different probabilistic models of NDE are connected within a unifying
framework. Using this framework, we derive insights into how these models
should be learned, calibrated, and applied. We investigate how the choice of
the model can affect the maintenance decisions taken on the basis of NDE
results. In addition, we analyze the impact of experimental design on the
performance of a given NDE system in a decision-making context.
| stat |
Synthetic learner: model-free inference on treatments over time | Understanding of the effect of a particular treatment or a policy pertains to
many areas of interest -- ranging from political economics, marketing to
health-care and personalized treatment studies. In this paper, we develop a
non-parametric, model-free test for detecting the effects of treatment over
time that extends widely used Synthetic Control tests. The test is built on
counterfactual predictions arising from many learning algorithms. In the
Neyman-Rubin potential outcome framework with possible carry-over effects, we
show that the proposed test is asymptotically consistent for stationary, beta
mixing processes. We do not assume that class of learners captures the correct
model necessarily. We also discuss estimates of the average treatment effect,
and we provide regret bounds on the predictive performance. To the best of our
knowledge, this is the first set of results that allow for example any Random
Forest to be useful for provably valid statistical inference in the Synthetic
Control setting. In experiments, we show that our Synthetic Learner is
substantially more powerful than classical methods based on Synthetic Control
or Difference-in-Differences, especially in the presence of non-linear outcome
models.
| stat |
Quaternion Recurrent Neural Networks | Recurrent neural networks (RNNs) are powerful architectures to model
sequential data, due to their capability to learn short and long-term
dependencies between the basic elements of a sequence. Nonetheless, popular
tasks such as speech or images recognition, involve multi-dimensional input
features that are characterized by strong internal dependencies between the
dimensions of the input vector. We propose a novel quaternion recurrent neural
network (QRNN), alongside with a quaternion long-short term memory neural
network (QLSTM), that take into account both the external relations and these
internal structural dependencies with the quaternion algebra. Similarly to
capsules, quaternions allow the QRNN to code internal dependencies by composing
and processing multidimensional features as single entities, while the
recurrent operation reveals correlations between the elements composing the
sequence. We show that both QRNN and QLSTM achieve better performances than RNN
and LSTM in a realistic application of automatic speech recognition. Finally,
we show that QRNN and QLSTM reduce by a maximum factor of 3.3x the number of
free parameters needed, compared to real-valued RNNs and LSTMs to reach better
results, leading to a more compact representation of the relevant information.
| stat |
Langevin Markov Chain Monte Carlo with stochastic gradients | Monte Carlo sampling techniques have broad applications in machine learning,
Bayesian posterior inference, and parameter estimation. Often the target
distribution takes the form of a product distribution over a dataset with a
large number of entries. For sampling schemes utilizing gradient information it
is cheaper for the derivative to be approximated using a random small subset of
the data, introducing extra noise into the system. We present a new
discretization scheme for underdamped Langevin dynamics when utilizing a
stochastic (noisy) gradient. This scheme is shown to bias computed averages to
second order in the stepsize while giving exact results in the special case of
sampling a Gaussian distribution with a normally distributed stochastic
gradient.
| stat |
The multilayer random dot product graph | We present a comprehensive extension of the latent position network model
known as the random dot product graph to accommodate multiple graphs -- both
undirected and directed -- which share a common subset of nodes, and propose a
method for jointly embedding the associated adjacency matrices, or submatrices
thereof, into a suitable latent space. Theoretical results concerning the
asymptotic behaviour of the node representations thus obtained are established,
showing that after the application of a linear transformation these converge
uniformly in the Euclidean norm to the latent positions with Gaussian error.
Within this framework, we present a generalisation of the stochastic block
model to a number of different multiple graph settings, and demonstrate the
effectiveness of our joint embedding method through several statistical
inference tasks in which we achieve comparable or better results than rival
spectral methods. Empirical improvements in link prediction over single graph
embeddings are exhibited in a cyber-security example.
| stat |
A Generalization Error Bound for Multi-class Domain Generalization | Domain generalization is the problem of assigning labels to an unlabeled data
set, given several similar data sets for which labels have been provided.
Despite considerable interest in this problem over the last decade, there has
been no theoretical analysis in the setting of multi-class classification. In
this work, we study a kernel-based learning algorithm and establish a
generalization error bound that scales logarithmically in the number of
classes, matching state-of-the-art bounds for multi-class classification in the
conventional learning setting. We also demonstrate empirically that the
proposed algorithm achieves significant performance gains compared to a pooling
strategy.
| stat |
Nonparametric goodness-of-fit testing for parametric covariate models in
pharmacometric analyses | The characterization of covariate effects on model parameters is a crucial
step during pharmacokinetic/pharmacodynamic analyses. While covariate selection
criteria have been studied extensively, the choice of the functional
relationship between covariates and parameters, however, has received much less
attention. Often, a simple particular class of covariate-to-parameter
relationships (linear, exponential, etc.) is chosen ad hoc or based on domain
knowledge, and a statistical evaluation is limited to the comparison of a small
number of such classes. Goodness-of-fit testing against a nonparametric
alternative provides a more rigorous approach to covariate model evaluation,
but no such test has been proposed so far. In this manuscript, we derive and
evaluate nonparametric goodness-of-fit tests for parametric covariate models,
the null hypothesis, against a kernelized Tikhonov regularized alternative,
transferring concepts from statistical learning to the pharmacological setting.
The approach is evaluated in a simulation study on the estimation of the
age-dependent maturation effect on the clearance of a monoclonal antibody.
Scenarios of varying data sparsity and residual error are considered. The
goodness-of-fit test correctly identified misspecified parametric models with
high power for relevant scenarios. The case study provides proof-of-concept of
the feasibility of the proposed approach, which is envisioned to be beneficial
for applications that lack well-founded covariate models.
| stat |
Conditional Monte Carlo revisited | Conditional Monte Carlo refers to sampling from the conditional distribution
of a random vector X given the value T(X) = t for a function T(X). Classical
conditional Monte Carlo methods were designed for estimating conditional
expectations of functions of X by sampling from unconditional distributions
obtained by certain weighting schemes. The basic ingredients were the use of
importance sampling and change of variables. In the present paper we
reformulate the problem by introducing an artificial parametric model,
representing the conditional distribution of X given T(X)=t within this new
model. The key is to provide the parameter of the artificial model by a
distribution. The approach is illustrated by several examples, which are
particularly chosen to illustrate conditional sampling in cases where such
sampling is not straightforward. A simulation study and an application to
goodness-of-fit testing of real data are also given.
| stat |
Orthogonal Statistical Inference for Multimodal Data Analysis | Multimodal imaging has transformed neuroscience research. While it presents
unprecedented opportunities, it also imposes serious challenges. Particularly,
it is difficult to combine the merits of interpretability attributed to a
simple association model and flexibility achieved by a highly adaptive
nonlinear model. In this article, we propose an orthogonal statistical
inferential framework, built upon the Neyman orthogonality and a form of
decomposition orthogonality, for multimodal data analysis. We target the
setting that naturally arises in almost all multimodal studies, where there is
a primary modality of interest, plus additional auxiliary modalities. We
successfully establish the root-$N$-consistency and asymptotic normality of the
estimated primary parameter, the semi-parametric estimation efficiency, and the
asymptotic honesty of the confidence interval of the predicted primary modality
effect. Our proposal enjoys, to a good extent, both model interpretability and
model flexibility. It is also considerably different from the existing
statistical methods for multimodal data integration, as well as the
orthogonality-based methods for high-dimensional inferences. We demonstrate the
efficacy of our method through both simulations and an application to a
multimodal neuroimaging study of Alzheimer's disease.
| stat |
Network Mediation Analysis Using Model-based Eigenvalue Decomposition | This paper proposes a new two-stage network mediation method based on the use
of a latent network approach -- model-based eigenvalue decomposition -- for
analyzing social network data with nodal covariates. In the decomposition stage
of the observed network, no assumption on the metric of the latent space
structure is required. In the mediation stage, the most important eigenvectors
of a network are used as mediators. This method further offers an innovative
way for controlling for the conditional covariates and it only considers the
information left in the network. We demonstrate this approach in a detailed
tutorial R code provided for four separate cases -- unconditional and
conditional model-based eigenvalue decompositions for either a continuous
outcome or a binary outcome -- to show its applicability to empirical network
data.
| stat |
A test for comparing conditional ROC curves with multidimensional
covariates | The comparison of Receiver Operating Characteristic (ROC) curves is
frequently used in the literature to compare the discriminatory capability of
different classification procedures based on diagnostic variables. The
performance of these variables can be sometimes influenced by the presence of
other covariates, and thus they should be taken into account when making the
comparison. A new non-parametric test is proposed here for testing the equality
of two or more dependent ROC curves conditioned to the value of a
multidimensional covariate. Projections are used for transforming the problem
into a one-dimensional approach easier to handle. Simulations are carried out
to study the practical performance of the new methodology. A real data set of
patients with Pleural Effusion is analysed to illustrate this procedure.
| stat |
Robust Wald-type test in GLM with random design based on minimum density
power divergence estimators | We consider the problem of robust inference under the generalized linear
model (GLM) with stochastic covariates. We derive the properties of the minimum
density power divergence estimator of the parameters in GLM with random design
and use this estimator to propose robust Wald-type tests for testing any
general composite null hypothesis about the GLM. The asymptotic and robustness
properties of the proposed tests are also examined for the GLM with random
design. Application of the proposed robust inference procedures to the popular
Poisson regression model for analyzing count data is discussed in detail both
theoretically and numerically through simulation studies and real data
examples.
| stat |
Heterogeneity-aware and communication-efficient distributed statistical
inference | In multicenter research, individual-level data are often protected against
sharing across sites. To overcome the barrier of data sharing, many distributed
algorithms, which only require sharing aggregated information, have been
developed. The existing distributed algorithms usually assume the data are
homogeneously distributed across sites. This assumption ignores the important
fact that the data collected at different sites may come from various
sub-populations and environments, which can lead to heterogeneity in the
distribution of the data. Ignoring the heterogeneity may lead to erroneous
statistical inference. In this paper, we propose distributed algorithms which
account for the heterogeneous distributions by allowing site-specific nuisance
parameters. The proposed methods extend the surrogate likelihood approach to
the heterogeneous setting by applying a novel density ratio tilting method to
the efficient score function. The proposed algorithms maintain the same
communication cost as the existing communication-efficient algorithms. We
establish a non-asymptotic risk bound for the proposed distributed estimator
and its limiting distribution in the two-index asymptotic setting which allows
both sample size per site and the number of sites to go to infinity. In
addition, we show that the asymptotic variance of the estimator attains the
Cram\'er-Rao lower bound when the number of sites is in rate smaller than the
sample size at each site. Finally, we use simulation studies and a real data
application to demonstrate the validity and feasibility of the proposed
methods.
| stat |
Causal Inference from Possibly Unbalanced Split-Plot Designs: A
Randomization-based Perspective | Split-plot designs find wide applicability in multifactor experiments with
randomization restrictions. Practical considerations often warrant the use of
unbalanced designs. This paper investigates randomization based causal
inference in split-plot designs that are possibly unbalanced. Extension of
ideas from the recently studied balanced case yields an expression for the
sampling variance of a treatment contrast estimator as well as a conservative
estimator of the sampling variance. However, the bias of this variance
estimator does not vanish even when the treatment effects are strictly
additive. A careful and involved matrix analysis is employed to overcome this
difficulty, resulting in a new variance estimator, which becomes unbiased under
milder conditions. A construction procedure that generates such an estimator
with minimax bias is proposed.
| stat |
NCVis: Noise Contrastive Approach for Scalable Visualization | Modern methods for data visualization via dimensionality reduction, such as
t-SNE, usually have performance issues that prohibit their application to large
amounts of high-dimensional data. In this work, we propose NCVis -- a
high-performance dimensionality reduction method built on a sound statistical
basis of noise contrastive estimation. We show that NCVis outperforms
state-of-the-art techniques in terms of speed while preserving the
representation quality of other methods. In particular, the proposed approach
successfully proceeds a large dataset of more than 1 million news headlines in
several minutes and presents the underlying structure in a human-readable way.
Moreover, it provides results consistent with classical methods like t-SNE on
more straightforward datasets like images of hand-written digits. We believe
that the broader usage of such software can significantly simplify the
large-scale data analysis and lower the entry barrier to this area.
| stat |
Risk-Efficient Bayesian Data Synthesis for Privacy Protection | Statistical agencies utilize models to synthesize respondent-level data for
release to the public for privacy protection. In this work, we efficiently
induce privacy protection into any Bayesian synthesis model by employing a
pseudo likelihood that exponentiates each likelihood contribution by an
observation record-indexed weight in [0, 1], defined to be inversely
proportional to the identification risk for that record. We start with the
marginal probability of identification risk for a record, which is composed as
the probability that the identity of the record may be disclosed. Our
application to the Consumer Expenditure Surveys (CE) of the U.S. Bureau of
Labor Statistics demonstrates that the marginally risk-adjusted synthesizer
provides an overall improved privacy protection; however, the identification
risks actually increase for some moderate-risk records after risk-adjusted
pseudo posterior estimation synthesis due to increased isolation after
weighting; a phenomenon we label "whack-a-mole". We proceed to construct a
weight for each record from a collection of pairwise identification risk
probabilities with other records, where each pairwise probability measures the
joint probability of re-identification of the pair of records, which mitigates
the whack-a-mole issue and produces a more efficient set of synthetic data with
lower risk and higher utility for the CE data.
| stat |
A Twin Neural Model for Uplift | Uplift is a particular case of conditional treatment effect modeling. Such
models deal with cause-and-effect inference for a specific factor, such as a
marketing intervention or a medical treatment. In practice, these models are
built on individual data from randomized clinical trials where the goal is to
partition the participants into heterogeneous groups depending on the uplift.
Most existing approaches are adaptations of random forests for the uplift case.
Several split criteria have been proposed in the literature, all relying on
maximizing heterogeneity. However, in practice, these approaches are prone to
overfitting. In this work, we bring a new vision to uplift modeling. We propose
a new loss function defined by leveraging a connection with the Bayesian
interpretation of the relative risk. Our solution is developed for a specific
twin neural network architecture allowing to jointly optimize the marginal
probabilities of success for treated and control individuals. We show that this
model is a generalization of the uplift logistic interaction model. We modify
the stochastic gradient descent algorithm to allow for structured sparse
solutions. This helps training our uplift models to a great extent. We show our
proposed method is competitive with the state-of-the-art in simulation setting
and on real data from large scale randomized experiments.
| stat |
Unbiased Implicit Variational Inference | We develop unbiased implicit variational inference (UIVI), a method that
expands the applicability of variational inference by defining an expressive
variational family. UIVI considers an implicit variational distribution
obtained in a hierarchical manner using a simple reparameterizable distribution
whose variational parameters are defined by arbitrarily flexible deep neural
networks. Unlike previous works, UIVI directly optimizes the evidence lower
bound (ELBO) rather than an approximation to the ELBO. We demonstrate UIVI on
several models, including Bayesian multinomial logistic regression and
variational autoencoders, and show that UIVI achieves both tighter ELBO and
better predictive performance than existing approaches at a similar
computational cost.
| stat |
Towards Deep Learning Models Resistant to Adversarial Attacks | Recent work has demonstrated that deep neural networks are vulnerable to
adversarial examples---inputs that are almost indistinguishable from natural
data and yet classified incorrectly by the network. In fact, some of the latest
findings suggest that the existence of adversarial attacks may be an inherent
weakness of deep learning models. To address this problem, we study the
adversarial robustness of neural networks through the lens of robust
optimization. This approach provides us with a broad and unifying view on much
of the prior work on this topic. Its principled nature also enables us to
identify methods for both training and attacking neural networks that are
reliable and, in a certain sense, universal. In particular, they specify a
concrete security guarantee that would protect against any adversary. These
methods let us train networks with significantly improved resistance to a wide
range of adversarial attacks. They also suggest the notion of security against
a first-order adversary as a natural and broad security guarantee. We believe
that robustness against such well-defined classes of adversaries is an
important stepping stone towards fully resistant deep learning models. Code and
pre-trained models are available at https://github.com/MadryLab/mnist_challenge
and https://github.com/MadryLab/cifar10_challenge.
| stat |
The Fuzzy ROC | The fuzzy ROC extends Receiver Operating Curve (ROC) visualization to the
situation where some data points, falling in an indeterminacy region, are not
classified. It addresses two challenges: definition of sensitivity and
specificity bounds under indeterminacy; and visual summarization of the large
number of possibilities arising from different choices of indeterminacy zones.
| stat |
AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data | We introduce AutoGluon-Tabular, an open-source AutoML framework that requires
only a single line of Python to train highly accurate machine learning models
on an unprocessed tabular dataset such as a CSV file. Unlike existing AutoML
frameworks that primarily focus on model/hyperparameter selection,
AutoGluon-Tabular succeeds by ensembling multiple models and stacking them in
multiple layers. Experiments reveal that our multi-layer combination of many
models offers better use of allocated training time than seeking out the best.
A second contribution is an extensive evaluation of public and commercial
AutoML platforms including TPOT, H2O, AutoWEKA, auto-sklearn, AutoGluon, and
Google AutoML Tables. Tests on a suite of 50 classification and regression
tasks from Kaggle and the OpenML AutoML Benchmark reveal that AutoGluon is
faster, more robust, and much more accurate. We find that AutoGluon often even
outperforms the best-in-hindsight combination of all of its competitors. In two
popular Kaggle competitions, AutoGluon beat 99% of the participating data
scientists after merely 4h of training on the raw data.
| stat |
Incorporating Posterior-Informed Approximation Errors into a
Hierarchical Framework to Facilitate Out-of-the-Box MCMC Sampling for
Geothermal Inverse Problems and Uncertainty Quantification | We consider geothermal inverse problems and uncertainty quantification from a
Bayesian perspective. Our main goal is to make standard, `out-of-the-box'
Markov chain Monte Carlo (MCMC) sampling more feasible for complex simulation
models by using suitable approximations. To do this, we first show how to pose
both the inverse and prediction problems in a hierarchical Bayesian framework.
We then show how to incorporate so-called posterior-informed model
approximation error into this hierarchical framework, using a modified form of
the Bayesian approximation error (BAE) approach. This enables the use of a
`coarse', approximate model in place of a finer, more expensive model, while
accounting for the additional uncertainty and potential bias that this can
introduce. Our method requires only simple probability modelling, a relatively
small number of fine model simulations, and only modifies the target posterior
-- any standard MCMC sampling algorithm can be used to sample the new
posterior. These corrections can also be used in methods that are not based on
MCMC sampling. We show that our approach can achieve significant computational
speed-ups on two geothermal test problems. We also demonstrate the dangers of
naively using coarse, approximate models in place of finer models, without
accounting for the induced approximation errors. The naive approach tends to
give overly confident and biased posteriors while incorporating BAE into our
hierarchical framework corrects for this while maintaining computational
efficiency and ease-of-use.
| stat |
Parallel Markov Chain Monte Carlo for Bayesian Hierarchical Models with
Big Data, in Two Stages | Due to the escalating growth of big data sets in recent years, new Bayesian
Markov chain Monte Carlo (MCMC) parallel computing methods have been developed.
These methods partition large data sets by observations into subsets. However,
for Bayesian nested hierarchical models, typically only a few parameters are
common for the full data set, with most parameters being group-specific. Thus,
parallel Bayesian MCMC methods that take into account the structure of the
model and split the full data set by groups rather than by observations are a
more natural approach for analysis. Here, we adapt and extend a recently
introduced two-stage Bayesian hierarchical modeling approach, and we partition
complete data sets by groups. In stage 1, the group-specific parameters are
estimated independently in parallel. The stage 1 posteriors are used as
proposal distributions in stage 2, where the target distribution is the full
model. Using three-level and four-level models, we show in both simulation and
real data studies that results of our method agree closely with the full data
analysis, with greatly increased MCMC efficiency and greatly reduced
computation times. The advantages of our method versus existing parallel MCMC
computing methods are also described.
| stat |
Random Machines Regression Approach: an ensemble support vector
regression model with free kernel choice | Machine learning techniques always aim to reduce the generalized prediction
error. In order to reduce it, ensemble methods present a good approach
combining several models that results in a greater forecasting capacity. The
Random Machines already have been demonstrated as strong technique, i.e: high
predictive power, to classification tasks, in this article we propose an
procedure to use the bagged-weighted support vector model to regression
problems. Simulation studies were realized over artificial datasets, and over
real data benchmarks. The results exhibited a good performance of Regression
Random Machines through lower generalization error without needing to choose
the best kernel function during tuning process.
| stat |
The Hellinger Correlation | In this paper, the defining properties of a valid measure of the dependence
between two random variables are reviewed and complemented with two original
ones, shown to be more fundamental than other usual postulates. While other
popular choices are proved to violate some of these requirements, a class of
dependence measures satisfying all of them is identified. One particular
measure, that we call the Hellinger correlation, appears as a natural choice
within that class due to both its theoretical and intuitive appeal. A simple
and efficient nonparametric estimator for that quantity is proposed. Synthetic
and real-data examples finally illustrate the descriptive ability of the
measure, which can also be used as test statistic for exact independence
testing.
| stat |
Handling Missingness Value on Jointly Measured Time-Course and
Time-to-event Data | Joint modeling technique is a recent advancement in effectively analyzing the
longitudinal history of patients with the occurrence of an event of interest
attached to it. This procedure is successfully implemented in biomarker studies
to examine parents with the occurrence of tumor. One of the typical problem
that influences the necessary inference is the presence of missing values in
the longitudinal responses as well as in covariates. The occurrence of
missingness is very common due to the dropout of patients from the study. This
article presents an effective and detailed way to handle the missing values in
the covariates and response variable. This study discusses the effect of
different multiple imputation techniques on the inferences of joint modeling
implemented on imputed datasets. A simulation study is carried out to replicate
the complex data structures and conveniently perform our analysis to show its
efficacy in terms of parameter estimation. This analysis is further illustrated
with the longitudinal and survival outcomes of biomarkers' study by assessing
proper codes in R programming language.
| stat |
Estimating uncertainty of earthquake rupture using Bayesian neural
network | Bayesian neural networks (BNN) are the probabilistic model that combines the
strengths of both neural network (NN) and stochastic processes. As a result,
BNN can combat overfitting and perform well in applications where data is
limited. Earthquake rupture study is such a problem where data is insufficient,
and scientists have to rely on many trial and error numerical or physical
models. Lack of resources and computational expenses, often, it becomes hard to
determine the reasons behind the earthquake rupture. In this work, a BNN has
been used (1) to combat the small data problem and (2) to find out the
parameter combinations responsible for earthquake rupture and (3) to estimate
the uncertainty associated with earthquake rupture. Two thousand rupture
simulations are used to train and test the model. A simple 2D rupture geometry
is considered where the fault has a Gaussian geometric heterogeneity at the
center, and eight parameters vary in each simulation. The test F1-score of BNN
(0.8334), which is 2.34% higher than plain NN score. Results show that the
parameters of rupture propagation have higher uncertainty than the rupture
arrest. Normal stresses play a vital role in determining rupture propagation
and are also the highest source of uncertainty, followed by the dynamic
friction coefficient. Shear stress has a moderate role, whereas the geometric
features such as the width and height of the fault are least significant and
uncertain.
| stat |
Optimizing the allocation of trials to sub-regions in multi-environment
crop variety testing | New crop varieties are extensively tested in multi-environment trials in
order to obtain a solid empirical basis for recommendations to farmers. When
the target population of environments is large and heterogeneous, a division
into sub-regions is often advantageous. When designing such trials, the
question arises how to allocate trials to the different subregions. We consider
a solution to this problem assuming a linear mixed model. We propose an
analytical approach for computation of optimal designs for best linear unbiased
prediction of genotype effects and pairwise linear contrasts and illustrate the
obtained results by a real data example from Indian nation-wide maize variety
trials. It is shown that, except in simple cases such as a compound symmetry
model, the optimal allocation depends on the variance-covariance structure for
genotypic effects nested within sub-regions.
| stat |
Gaussian Markov Random Fields versus Linear Mixed Models for
satellite-based PM2.5 assessment: Evidence from the Northeastern USA | Studying the effects of air-pollution on health is a key area in
environmental epidemiology. An accurate estimation of air-pollution effects
requires spatio-temporally resolved datasets of air-pollution, especially, Fine
Particulate Matter (PM). Satellite-based technology has greatly enhanced the
ability to provide PM assessments in locations where direct measurement is
impossible.
Indirect PM measurement is a statistical prediction problem. The
spatio-temporal statistical literature offer various predictive models:
Gaussian Random Fields (GRF) and Linear Mixed Models (LMM), in particular. GRF
emphasize the spatio-temporal structure in the data, but are computationally
demanding to fit. LMMs are computationally easier to fit, but require some
tampering to deal with space and time.
Recent advances in the spatio-temporal statistical literature propose to
alleviate the computation burden of GRFs by approximating them with Gaussian
Markov Random Fields (GMRFs). Since LMMs and GMRFs are both computationally
feasible, the question arises: which is statistically better? We show that
despite the great popularity of LMMs in environmental monitoring and pollution
assessment, LMMs are statistically inferior to GMRF for measuring PM in the
Northeastern USA.
| stat |
Keyword-based Topic Modeling and Keyword Selection | Certain type of documents such as tweets are collected by specifying a set of
keywords. As topics of interest change with time it is beneficial to adjust
keywords dynamically. The challenge is that these need to be specified ahead of
knowing the forthcoming documents and the underlying topics. The future topics
should mimic past topics of interest yet there should be some novelty in them.
We develop a keyword-based topic model that dynamically selects a subset of
keywords to be used to collect future documents. The generative process first
selects keywords and then the underlying documents based on the specified
keywords. The model is trained by using a variational lower bound and
stochastic gradient optimization. The inference consists of finding a subset of
keywords where given a subset the model predicts the underlying topic-word
matrix for the unknown forthcoming documents. We compare the keyword topic
model against a benchmark model using viral predictions of tweets combined with
a topic model. The keyword-based topic model outperforms this sophisticated
baseline model by 67%.
| stat |
Decentralized learning with budgeted network load using Gaussian copulas
and classifier ensembles | We examine a network of learners which address the same classification task
but must learn from different data sets. The learners cannot share data but
instead share their models. Models are shared only one time so as to preserve
the network load. We introduce DELCO (standing for Decentralized Ensemble
Learning with COpulas), a new approach allowing to aggregate the predictions of
the classifiers trained by each learner. The proposed method aggregates the
base classifiers using a probabilistic model relying on Gaussian copulas.
Experiments on logistic regressor ensembles demonstrate competing accuracy and
increased robustness in case of dependent classifiers. A companion python
implementation can be downloaded at https://github.com/john-klein/DELCO
| stat |
Corruption-Tolerant Gaussian Process Bandit Optimization | We consider the problem of optimizing an unknown (typically non-convex)
function with a bounded norm in some Reproducing Kernel Hilbert Space (RKHS),
based on noisy bandit feedback. We consider a novel variant of this problem in
which the point evaluations are not only corrupted by random noise, but also
adversarial corruptions. We introduce an algorithm Fast-Slow GP-UCB based on
Gaussian process methods, randomized selection between two instances labeled
"fast" (but non-robust) and "slow" (but robust), enlarged confidence bounds,
and the principle of optimism under uncertainty. We present a novel theoretical
analysis upper bounding the cumulative regret in terms of the corruption level,
the time horizon, and the underlying kernel, and we argue that certain
dependencies cannot be improved. We observe that distinct algorithmic ideas are
required depending on whether one is required to perform well in both the
corrupted and non-corrupted settings, and whether the corruption level is known
or not.
| stat |
Cost-sensitive Multi-class AdaBoost for Understanding Driving Behavior
with Telematics | Powered with telematics technology, insurers can now capture a wide range of
data, such as distance traveled, how drivers brake, accelerate or make turns,
and travel frequency each day of the week, to better decode driver's behavior.
Such additional information helps insurers improve risk assessments for
usage-based insurance (UBI), an increasingly popular industry innovation. In
this article, we explore how to integrate telematics information to better
predict claims frequency. For motor insurance during a policy year, we
typically observe a large proportion of drivers with zero claims, a less
proportion with exactly one claim, and far lesser with two or more claims. We
introduce the use of a cost-sensitive multi-class adaptive boosting (AdaBoost)
algorithm, which we call SAMME.C2, to handle such imbalances. To calibrate
SAMME.C2 algorithm, we use empirical data collected from a telematics program
in Canada and we find improved assessment of driving behavior with telematics
relative to traditional risk variables. We demonstrate our algorithm can
outperform other models that can handle class imbalances: SAMME, SAMME with
SMOTE, RUSBoost, and SMOTEBoost. The sampled data on telematics were
observations during 2013-2016 for which 50,301 are used for training and
another 21,574 for testing. Broadly speaking, the additional information
derived from vehicle telematics helps refine risk classification of drivers of
UBI.
| stat |
On Disentanglement in Gaussian Process Variational Autoencoders | Complex multivariate time series arise in many fields, ranging from computer
vision to robotics or medicine. Often we are interested in the independent
underlying factors that give rise to the high-dimensional data we are
observing. While many models have been introduced to learn such disentangled
representations, only few attempt to explicitly exploit the structure of
sequential data. We investigate the disentanglement properties of Gaussian
process variational autoencoders, a class of models recently introduced that
have been successful in different tasks on time series data. Our model exploits
the temporal structure of the data by modeling each latent channel with a GP
prior and employing a structured variational distribution that can capture
dependencies in time. We demonstrate the competitiveness of our approach
against state-of-the-art unsupervised and weakly-supervised disentanglement
methods on a benchmark task. Moreover, we provide evidence that we can learn
meaningful disentangled representations on real-world medical time series data.
| stat |
Anytime Parallel Tempering | Developing efficient MCMC algorithms is indispensable in Bayesian inference.
In parallel tempering, multiple interacting MCMC chains run to more efficiently
explore the state space and improve performance. The multiple chains advance
independently through local moves, and the performance enhancement steps are
exchange moves, where the chains pause to exchange their current sample amongst
each other. To accelerate the independent local moves, they may be performed
simultaneously on multiple processors. Another problem is then encountered:
depending on the MCMC implementation and inference problem, local moves can
take a varying and random amount of time to complete. There may also be
infrastructure-induced variations, such as competing jobs on the same
processors, which arises in cloud computing. Before exchanges can occur, all
chains must complete the local moves they are engaged in to avoid introducing a
potentially substantial bias (Proposition 2.1). To solve this issue of randomly
varying local move completion times in multi-processor parallel tempering, we
adopt the Anytime Monte Carlo framework of Murray et al. (2016): we impose
real-time deadlines on the parallel local moves and perform exchanges at these
deadlines without any processor idling. We show our methodology for exchanges
at real-time deadlines does not introduce a bias and leads to significant
performance enhancements over the na\"ive approach of idling until every
processor's local moves complete. The methodology is then applied in an ABC
setting, where an Anytime ABC parallel tempering algorithm is derived for the
difficult task of estimating the parameters of a Lotka-Volterra predator-prey
model, and similar efficiency enhancements are observed.
| stat |
A Kernel-based Consensual Aggregation for Regression | In this article, we introduce a kernel-based consensual aggregation method
for regression problems. We aim to flexibly combine individual regression
estimators $r_1, r_2, \ldots, r_M$ using a weighted average where the weights
are defined based on some kernel function to build a target prediction. This
work extends the context of Biau et al. (2016) to a more general kernel-based
framework. We show that this more general configuration also inherits the
consistency of the basic consistent estimators. Moreover, an optimization
method based on gradient descent algorithm is proposed to efficiently and
rapidly estimate the key parameter of the strategy. The numerical experiments
carried out on several simulated and real datasets are also provided to
illustrate the speed-up of gradient descent algorithm in estimating the key
parameter and the improvement of overall performance of the method with the
introduction of smoother kernel functions.
| stat |
Invariant Risk Minimization | We introduce Invariant Risk Minimization (IRM), a learning paradigm to
estimate invariant correlations across multiple training distributions. To
achieve this goal, IRM learns a data representation such that the optimal
classifier, on top of that data representation, matches for all training
distributions. Through theory and experiments, we show how the invariances
learned by IRM relate to the causal structures governing the data and enable
out-of-distribution generalization.
| stat |
Hamiltonian Assisted Metropolis Sampling | Various Markov chain Monte Carlo (MCMC) methods are studied to improve upon
random walk Metropolis sampling, for simulation from complex distributions.
Examples include Metropolis-adjusted Langevin algorithms, Hamiltonian Monte
Carlo, and other recent algorithms related to underdamped Langevin dynamics. We
propose a broad class of irreversible sampling algorithms, called Hamiltonian
assisted Metropolis sampling (HAMS), and develop two specific algorithms with
appropriate tuning and preconditioning strategies. Our HAMS algorithms are
designed to achieve two distinctive properties, while using an augmented target
density with momentum as an auxiliary variable. One is generalized detailed
balance, which induces an irreversible exploration of the target. The other is
a rejection-free property, which allows our algorithms to perform
satisfactorily with relatively large step sizes. Furthermore, we formulate a
framework of generalized Metropolis--Hastings sampling, which not only
highlights our construction of HAMS at a more abstract level, but also
facilitates possible further development of irreversible MCMC algorithms. We
present several numerical experiments, where the proposed algorithms are found
to consistently yield superior results among existing ones.
| stat |
Dynamic principal component regression: Application to age-specific
mortality forecasting | In areas of application, including actuarial science and demography, it is
increasingly common to consider a time series of curves; an example of this is
age-specific mortality rates observed over a period of years. Given that age
can be treated as a discrete or continuous variable, a dimension reduction
technique, such as principal component analysis, is often implemented. However,
in the presence of moderate to strong temporal dependence, static principal
component analysis commonly used for analyzing independent and identically
distributed data may not be adequate. As an alternative, we consider a
\textit{dynamic} principal component approach to model temporal dependence in a
time series of curves. Inspired by Brillinger's (1974) theory of dynamic
principal components, we introduce a dynamic principal component analysis,
which is based on eigen-decomposition of estimated long-run covariance. Through
a series of empirical applications, we demonstrate the potential improvement of
one-year-ahead point and interval forecast accuracies that the dynamic
principal component regression entails when compared with the static
counterpart.
| stat |
Automatic, Dynamic, and Nearly Optimal Learning Rate Specification by
Local Quadratic Approximation | In deep learning tasks, the learning rate determines the update step size in
each iteration, which plays a critical role in gradient-based optimization.
However, the determination of the appropriate learning rate in practice
typically replies on subjective judgement. In this work, we propose a novel
optimization method based on local quadratic approximation (LQA). In each
update step, given the gradient direction, we locally approximate the loss
function by a standard quadratic function of the learning rate. Then, we
propose an approximation step to obtain a nearly optimal learning rate in a
computationally efficient way. The proposed LQA method has three important
features. First, the learning rate is automatically determined in each update
step. Second, it is dynamically adjusted according to the current loss function
value and the parameter estimates. Third, with the gradient direction fixed,
the proposed method leads to nearly the greatest reduction in terms of the loss
function. Extensive experiments have been conducted to prove the strengths of
the proposed LQA method.
| stat |
Implicit Priors for Knowledge Sharing in Bayesian Neural Networks | Bayesian interpretations of neural network have a long history, dating back
to early work in the 1990's and have recently regained attention because of
their desirable properties like uncertainty estimation, model robustness and
regularisation. We want to discuss here the application of Bayesian models to
knowledge sharing between neural networks. Knowledge sharing comes in different
facets, such as transfer learning, model distillation and shared embeddings.
All of these tasks have in common that learned "features" ought to be shared
across different networks. Theoretically rooted in the concepts of Bayesian
neural networks this work has widespread application to general deep learning.
| stat |
Simultaneous comparisons of treatments versus control (Dunnett-type
tests) for location-scale alternatives | Commonly, the comparisons of treatment groups versus a control is performed
for location effects only where possible scale effects are considered as
disturbing. Sometimes scale effects are also relevant, as a kind of early
indicator for changes. Here several approaches for Dunnett-type tests for
location or scale effects are proposed and compared by a simulation study. Two
real data examples are analysed accordingly and the related R-code is available
in the Appendix.
| stat |
Bayes' Theorem under Conditional Independence | In this article we provide a substantial discussion on the statistical
concept of conditional independence, which is not routinely mentioned in most
elementary statistics and mathematical statistics textbooks. Under the
assumption of conditional independence, an extended version of Bayes' Theorem
is then proposed with illustrations from both hypothetical and real-world
examples of disease diagnosis.
| stat |
Pathological spectra of the Fisher information metric and its variants
in deep neural networks | The Fisher information matrix (FIM) plays an essential role in statistics and
machine learning as a Riemannian metric tensor or a component of the Hessian
matrix of loss functions. Focusing on the FIM and its variants in deep neural
networks (DNNs), we reveal their characteristic scale dependence on the network
width, depth and sample size when the network has random weights and is
sufficiently wide. This study covers two widely-used FIMs for regression with
linear output and for classification with softmax output. Both FIMs
asymptotically show pathological eigenvalue spectra in the sense that a small
number of eigenvalues become large outliers depending the width or sample size
while the others are much smaller. It implies that the local shape of the
parameter space or loss landscape is very sharp in a few specific directions
while almost flat in the other directions. In particular, the softmax output
disperses the outliers and makes a tail of the eigenvalue density spread from
the bulk. We also show that pathological spectra appear in other variants of
FIMs: one is the neural tangent kernel; another is a metric for the input
signal and feature space that arises from feedforward signal propagation. Thus,
we provide a unified perspective on the FIM and its variants that will lead to
more quantitative understanding of learning in large-scale DNNs.
| stat |
Multi-resolution Super Learner for Voxel-wise Classification of Prostate
Cancer Using Multi-parametric MRI | While current research has shown the importance of Multi-parametric MRI
(mpMRI) in diagnosing prostate cancer (PCa), further investigation is needed
for how to incorporate the specific structures of the mpMRI data, such as the
regional heterogeneity and between-voxel correlation within a subject. This
paper proposes a machine learning-based method for improved voxel-wise PCa
classification by taking into account the unique structures of the data. We
propose a multi-resolution modeling approach to account for regional
heterogeneity, where base learners trained locally at multiple resolutions are
combined using the super learner, and account for between-voxel correlation by
efficient spatial Gaussian kernel smoothing. The method is flexible in that the
super learner framework allows implementation of any classifier as the base
learner, and can be easily extended to classifying cancer into more
sub-categories. We describe detailed classification algorithm for the binary
PCa status, as well as the ordinal clinical significance of PCa for which a
weighted likelihood approach is implemented to enhance the detection of the
less prevalent cancer categories. We illustrate the advantages of the proposed
approach over conventional modeling and machine learning approaches through
simulations and application to in vivo data.
| stat |
Normalization of regressor excitation as a part of dynamic regressor
extension and mixing procedure | The method of excitation normalization of the regressor, which is used in the
estimation loop to solve the plant identification problem, is proposed. It is
based on the dynamic regressor extension and mixing procedure. Its application
allows to obtain the same upper bound of the parameter identification error for
the scalar regressors with different excitation level, using a constant value
of the adaptation rate for all of them. This fact is a significant advantage
from the practical point of view. Comparison of the developed method with the
known one of the regressor amplitude normalization is conducted. It is shown
that the classical approach does not have the above-stated property. To
validate the theoretical conclusions made, the results of the comparative
mathematical modeling of three loops are presented: 1) the classical gradient
one, 2) the one with the normalization of the regressor amplitude, 3) the
proposed one with the normalization of the regressor excitation.
| stat |
Preventing Failures Due to Dataset Shift: Learning Predictive Models
That Transport | Classical supervised learning produces unreliable models when training and
target distributions differ, with most existing solutions requiring samples
from the target domain. We propose a proactive approach which learns a
relationship in the training domain that will generalize to the target domain
by incorporating prior knowledge of aspects of the data generating process that
are expected to differ as expressed in a causal selection diagram.
Specifically, we remove variables generated by unstable mechanisms from the
joint factorization to yield the Surgery Estimator---an interventional
distribution that is invariant to the differences across environments. We prove
that the surgery estimator finds stable relationships in strictly more
scenarios than previous approaches which only consider conditional
relationships, and demonstrate this in simulated experiments. We also evaluate
on real world data for which the true causal diagram is unknown, performing
competitively against entirely data-driven approaches.
| stat |
Spatio-Temporal Change of Support Modeling with R | Spatio-temporal change of support methods are designed for statistical
analysis on spatial and temporal domains which can differ from those of the
observed data. Previous work introduced a parsimonious class of Bayesian
hierarchical spatio-temporal models, which we refer to as STCOS, for the case
of Gaussian outcomes. Application of STCOS methodology from this literature
requires a level of proficiency with spatio-temporal methods and statistical
computing which may be a hurdle for potential users. The present work seeks to
bridge this gap by guiding readers through STCOS computations. We focus on the
R computing environment because of its popularity, free availability, and high
quality contributed packages. The stcos package is introduced to facilitate
computations for the STCOS model. A motivating application is the American
Community Survey (ACS), an ongoing survey administered by the U.S. Census
Bureau that measures key socioeconomic and demographic variables for various
populations in the United States. The STCOS methodology offers a principled
approach to compute model-based estimates and associated measures of
uncertainty for ACS variables on customized geographies and/or time periods. We
present a detailed case study with ACS data as a guide for change of support
analysis in R, and as a foundation which can be customized to other
applications.
| stat |
Assessing the accuracy of individual link with varying block sizes and
cut-off values using MaCSim approach | Record linkage is the process of matching together records from different
data sources that belong to the same entity. Record linkage is increasingly
being used by many organizations including statistical, health, government etc.
to link administrative, survey, and other files to create a robust file for
more comprehensive analysis. Therefore, it becomes necessary to assess the
ability of a linking method to achieve high accuracy or compare between methods
with respect to accuracy. In this paper, we evaluate the accuracy of individual
link using varying block sizes and different cut-off values by utilizing a
Markov Chain based Monte Carlo simulation approach (MaCSim). MaCSim utilizes
two linked files to create an agreement matrix. The agreement matrix is
simulated to generate re-sampled versions of the agreement matrix. A defined
linking method is used in each simulation to link the files and the accuracy of
the linking method is assessed. The aim of this paper is to facilitate optimal
choice of block size and cut-off value to achieve high accuracy in terms of
minimizing average False Discovery Rate and False Negative Rate. The analyses
have been performed using a synthetic dataset provided by the Australian Bureau
of Statistics (ABS) and indicated promising results.
| stat |
An empirical analysis of the spatial variability of fuel prices in the
United States | In this paper, we use a newly constructed dataset to study the geographic
distribution of fuel price across the US at a very high resolution. We study
the influence of socio-economic variables through different and complementary
statistical methods. We highlight an optimal spatial range roughly
corresponding to stationarity scale, and significant influence of variables
such as median income, wage with a non-simple spatial behavior that confirms
the importance of geographical particularities. On the other hand, multi-level
modeling reveals a strong influence of the state in the level of price but also
of some local characteristics including population density. Through the
combination of such methods, we unveil the superposition of a governance
process with a local socio-economical spatial process. The influence of
population density on prices is furthermore consistent with a minimal
theoretical model of competition between gas stations, that we introduce and
solve numerically. We discuss developments and applications, including the
elaboration of locally parametrized car-regulation policies.
| stat |
Generalized Doubly Reparameterized Gradient Estimators | Efficient low-variance gradient estimation enabled by the reparameterization
trick (RT) has been essential to the success of variational autoencoders.
Doubly-reparameterized gradients (DReGs) improve on the RT for multi-sample
variational bounds by applying reparameterization a second time for an
additional reduction in variance. Here, we develop two generalizations of the
DReGs estimator and show that they can be used to train conditional and
hierarchical VAEs on image modelling tasks more effectively. We first extend
the estimator to hierarchical models with several stochastic layers by showing
how to treat additional score function terms due to the hierarchical
variational posterior. We then generalize DReGs to score functions of arbitrary
distributions instead of just those of the sampling distribution, which makes
the estimator applicable to the parameters of the prior in addition to those of
the posterior.
| stat |
Scanner data in inflation measurement: from raw data to price indices | Scanner data offer new opportunities for CPI or HICP calculation. They can be
obtained from a~wide variety of~retailers (supermarkets, home electronics,
Internet shops, etc.) and provide information at the level of~the barcode. One
of~advantages of~using scanner data is the fact that they contain complete
transaction information, i.e. prices and quantities for every sold item. To use
scanner data, it must be carefully processed. After clearing data and unifying
product names, products should be carefully classified (e.g. into COICOP 5 or
below), matched, filtered and aggregated. These procedures often require
creating new IT or writing custom scripts (R, Python, Mathematica, SAS,
others). One of~new challenges connected with scanner data is the appropriate
choice of~the index formula. In this article we present a~proposal for the
implementation of~individual stages of~handling scanner data. We also point out
potential problems during scanner data processing and their solutions. Finally,
we compare a~large number of~price index methods based on real scanner datasets
and we verify their sensitivity on adopted data filtering and aggregating
methods.
| stat |
Testing For a Parametric Baseline-Intensity in Dynamic Interaction
Networks | In statistical network analysis it is common to observe so called interaction
data. Such data is characterized by the actors who form the vertices of a
network. These are able to interact with each other along the edges of the
network. One usually assumes that the edges in the network are randomly formed
and dissolved over the observation horizon. In addition covariates are observed
and the interest is to model the impact of the covariates on the interactions.
In this paper we develop a framework to test if a non-parametric form of the
baseline intensity allows for more flexibility than a baseline which is
parametrically dependent on system-wide covariates (i.e. covariates which take
the same value for all individuals, e.g. time). This allows to test if certain
seasonality effects can be explained by simple covariates like the time. The
procedure is applied to modeling the baseline intensity in a bike-sharing
network by using weather and time information.
| stat |
Bayesian Inference of a Finite Population Mean Under Length-Biased
Sampling | We present a robust Bayesian method to analyze forestry data when samples are
selected with probability proportional to length from a finite population of
unknown size. Specifically, we use Bayesian predictive inference to estimate
the finite population mean of shrub widths in a limestone quarry dominated by
re-growth of mountain mahogany. The data on shrub widths are collected using
transect sampling and it is assumed that the probability that a shrub is
selected is proportional to its width; this is length-biased sampling. In this
type of sampling, the population size is also unknown and this creates an
additional challenge. The quantity of interest is average finite population
shrub width and the total shrub area of the quarry can be estimated. Our method
is assisted by using the three-parameter generalized gamma distribution,
thereby robustifying our procedure against a possible model failure. Using
conditional predictive ordinates, we show that the model, which accommodates
length bias, performs better than the model that does not. In the Bayesian
computation, we overcome a technical problem associated with Gibbs sampling by
using a random sampler.
| stat |
StateSpaceModels.jl: a Julia Package for Time-Series Analysis in a
State-Space Framework | StateSpaceModels.jl is an open-source Julia package for modeling, forecasting
and simulating time series in a state-space framework. The package represents a
straightforward tool that can be useful for a wide range of applications that
deal with time series. In addition, it contains features that are not present
in related commercial software, such as Monte Carlo simulation and the
possibility of setting any user-defined linear model.
| stat |
Robust Generalised Bayesian Inference for Intractable Likelihoods | Generalised Bayesian inference updates prior beliefs using a loss function,
rather than a likelihood, and can therefore be used to confer robustness
against possible misspecification of the likelihood. Here we consider
generalised Bayesian inference with a Stein discrepancy as a loss function,
motivated by applications in which the likelihood contains an intractable
normalisation constant. In this context, the Stein discrepancy circumvents
evaluation of the normalisation constant and produces generalised posteriors
that are either closed form or accessible using standard Markov chain Monte
Carlo. On a theoretical level, we show consistency, asymptotic normality, and
bias-robustness of the generalised posterior, highlighting how these properties
are impacted by the choice of Stein discrepancy. Then, we provide numerical
experiments on a range of intractable distributions, including applications to
kernel-based exponential family models and non-Gaussian graphical models.
| stat |
Particulate Matter Exposure and Lung Cancer: A Review of two
Meta-Analysis Studies | The current regulatory paradigm is that PM2.5, over time causes lung cancer.
This claim is based on cohort studies and meta-analysis that use cohort studies
as their base studies. There is a need to evaluate the reliability of this
causal claim. Our idea is to examine the base studies with respect to multiple
testing and multiple modeling and to look closer at the meta-analysis using
p-value plots. For two meta-analysis we investigated, some extremely small
p-values were observed in some of the base studies, which we think are due to a
combination of bias and small standard errors. The p-value plot for one
meta-analysis indicates no effect. For the other meta-analysis, we note the
p-value plot is consistent with a two-component mixture. Small p-values might
be real or due to some combination of p-hacking, publication bias, covariate
problems, etc. The large p-values could indicate no real effect, or be wrong
due to low power, missing covariates, etc. We conclude that the results are
ambiguous at best. These meta-analyses do not establish that PM2.5 is causal of
lung tumors.
| stat |
SPONGE: A generalized eigenproblem for clustering signed networks | We introduce a principled and theoretically sound spectral method for $k$-way
clustering in signed graphs, where the affinity measure between nodes takes
either positive or negative values. Our approach is motivated by social balance
theory, where the task of clustering aims to decompose the network into
disjoint groups, such that individuals within the same group are connected by
as many positive edges as possible, while individuals from different groups are
connected by as many negative edges as possible. Our algorithm relies on a
generalized eigenproblem formulation inspired by recent work on constrained
clustering. We provide theoretical guarantees for our approach in the setting
of a signed stochastic block model, by leveraging tools from matrix
perturbation theory and random matrix theory. An extensive set of numerical
experiments on both synthetic and real data shows that our approach compares
favorably with state-of-the-art methods for signed clustering, especially for
large number of clusters and sparse measurement graphs.
| stat |
Inferring UK COVID-19 fatal infection trajectories from daily mortality
data: were infections already in decline before the UK lockdowns? | The number of new infections per day is a key quantity for effective epidemic
management. It can be estimated relatively directly by testing of random
population samples. Without such direct epidemiological measurement, other
approaches are required to infer whether the number of new cases is likely to
be increasing or decreasing: for example, estimating the pathogen effective
reproduction number, R, using data gathered from the clinical response to the
disease. For Covid-19 (SARS-CoV-2) such R estimation is heavily dependent on
modelling assumptions, because the available clinical case data are
opportunistic observational data subject to severe temporal confounding. Given
this difficulty it is useful to retrospectively reconstruct the time course of
infections from the least compromised available data, using minimal prior
assumptions. A Bayesian inverse problem approach applied to UK data on first
wave Covid-19 deaths and the disease duration distribution suggests that fatal
infections were in decline before full UK lockdown (24 March 2020), and that
fatal infections in Sweden started to decline only a day or two later. An
analysis of UK data using the model of Flaxman et al. (2020, Nature 584) gives
the same result under relaxation of its prior assumptions on R, suggesting an
enhanced role for non pharmaceutical interventions (NPIs) short of full lock
down in the UK context. Similar patterns appear to have occurred in the
subsequent two lockdowns.
| stat |
Sample size planning for pilot studies | Pilot studies are often the first step of experimental research. It is
usually on a smaller scale and the results can inform intervention development,
study feasibility and how the study implementation will play out, if such a
larger main study is undertaken. This paper illustrates the relationship
between pilot study sample size and the performance study design of main
studies. We present two simple sample size calculation methods to ensure
adequate study planning for main studies. We use numerical examples and
simulations to demonstrate the use and performance of proposed methods.
Practical heuristic guidelines are provided based on the results.
| stat |
On spike-and-slab priors for Bayesian equation discovery of nonlinear
dynamical systems via sparse linear regression | This paper presents the use of spike-and-slab (SS) priors for discovering
governing differential equations of motion of nonlinear structural dynamic
systems. The problem of discovering governing equations is cast as that of
selecting relevant variables from a predetermined dictionary of basis variables
and solved via sparse Bayesian linear regression. The SS priors, which belong
to a class of discrete-mixture priors and are known for their strong
sparsifying (or shrinkage) properties, are employed to induce sparse solutions
and select relevant variables. Three different variants of SS priors are
explored for performing Bayesian equation discovery. As the posteriors with SS
priors are analytically intractable, a Markov chain Monte Carlo (MCMC)-based
Gibbs sampler is employed for drawing posterior samples of the model
parameters; the posterior samples are used for variable selection and parameter
estimation in equation discovery. The proposed algorithm has been applied to
four systems of engineering interest, which include a baseline linear system,
and systems with cubic stiffness, quadratic viscous damping, and Coulomb
damping. The results demonstrate the effectiveness of the SS priors in
identifying the presence and type of nonlinearity in the system. Additionally,
comparisons with the Relevance Vector Machine (RVM) - that uses a Student's-t
prior - indicate that the SS priors can achieve better model selection
consistency, reduce false discoveries, and derive models that have superior
predictive accuracy. Finally, the Silverbox experimental benchmark is used to
validate the proposed methodology.
| stat |
On Kernel Derivative Approximation with Random Fourier Features | Random Fourier features (RFF) represent one of the most popular and
wide-spread techniques in machine learning to scale up kernel algorithms.
Despite the numerous successful applications of RFFs, unfortunately, quite
little is understood theoretically on their optimality and limitations of their
performance. Only recently, precise statistical-computational trade-offs have
been established for RFFs in the approximation of kernel values, kernel ridge
regression, kernel PCA and SVM classification. Our goal is to spark the
investigation of optimality of RFF-based approximations in tasks involving not
only function values but derivatives, which naturally lead to optimization
problems with kernel derivatives. Particularly, in this paper, we focus on the
approximation quality of RFFs for kernel derivatives and prove that the
existing finite-sample guarantees can be improved exponentially in terms of the
domain where they hold, using recent tools from unbounded empirical process
theory. Our result implies that the same approximation guarantee is attainable
for kernel derivatives using RFF as achieved for kernel values.
| stat |
Supervised clustering of high dimensional data using regularized mixture
modeling | Identifying relationships between molecular variations and their clinical
presentations has been challenged by the heterogeneous causes of a disease. It
is imperative to unveil the relationship between the high dimensional molecular
manifestations and the clinical presentations, while taking into account the
possible heterogeneity of the study subjects. We proposed a novel supervised
clustering algorithm using penalized mixture regression model, called CSMR, to
deal with the challenges in studying the heterogeneous relationships between
high dimensional molecular features to a phenotype. The algorithm was adapted
from the classification expectation maximization algorithm, which offers a
novel supervised solution to the clustering problem, with substantial
improvement on both the computational efficiency and biological
interpretability. Experimental evaluation on simulated benchmark datasets
demonstrated that the CSMR can accurately identify the subspaces on which
subset of features are explanatory to the response variables, and it
outperformed the baseline methods. Application of CSMR on a drug sensitivity
dataset again demonstrated the superior performance of CSMR over the others,
where CSMR is powerful in recapitulating the distinct subgroups hidden in the
pool of cell lines with regards to their coping mechanisms to different drugs.
CSMR represents a big data analysis tool with the potential to resolve the
complexity of translating the clinical manifestations of the disease to the
real causes underpinning it. We believe that it will bring new understanding to
the molecular basis of a disease, and could be of special relevance in the
growing field of personalized medicine.
| stat |
Universal Boosting Variational Inference | Boosting variational inference (BVI) approximates an intractable probability
density by iteratively building up a mixture of simple component distributions
one at a time, using techniques from sparse convex optimization to provide both
computational scalability and approximation error guarantees. But the
guarantees have strong conditions that do not often hold in practice, resulting
in degenerate component optimization problems; and we show that the ad-hoc
regularization used to prevent degeneracy in practice can cause BVI to fail in
unintuitive ways. We thus develop universal boosting variational inference
(UBVI), a BVI scheme that exploits the simple geometry of probability densities
under the Hellinger metric to prevent the degeneracy of other gradient-based
BVI methods, avoid difficult joint optimizations of both component and weight,
and simplify fully-corrective weight optimizations. We show that for any target
density and any mixture component family, the output of UBVI converges to the
best possible approximation in the mixture family, even when the mixture family
is misspecified. We develop a scalable implementation based on exponential
family mixture components and standard stochastic optimization techniques.
Finally, we discuss statistical benefits of the Hellinger distance as a
variational objective through bounds on posterior probability, moment, and
importance sampling errors. Experiments on multiple datasets and models show
that UBVI provides reliable, accurate posterior approximations.
| stat |
Evaluating real-time probabilistic forecasts with application to
National Basketball Association outcome prediction | Motivated by the goal of evaluating real-time forecasts of home team win
probabilities in the National Basketball Association, we develop new tools for
measuring the quality of continuously updated probabilistic forecasts. This
includes introducing calibration surface plots, and simple graphical summaries
of them, to evaluate at a glance whether a given continuously updated
probability forecasting method is well-calibrated, as well as developing
statistical tests and graphical tools to evaluate the skill, or relative
performance, of two competing continuously updated forecasting methods. These
tools are studied by means of a Monte Carlo simulation study of simulated
basketball games, and demonstrated in an application to evaluate the
continuously updated forecasts published by the United States-based
multinational sports network ESPN on its principle webpage {\tt espn.com}. This
application lends statistical evidence that the forecasts published there are
well-calibrated, and exhibit improved skill over several na\"ive models, but do
not demonstrate significantly improved skill over simple logistic regression
models based solely on a measurement of each teams' relative strength, and the
evolving score difference throughout the game.
| stat |
Design aspects of COVID-19 treatment trials: Improving probability and
time of favourable events | As a reaction to the pandemic of the severe acute respiratory syndrome
coronavirus 2 (SARS-CoV-2), a multitude of clinical trials for the treatment of
SARS-CoV-2 or the resulting corona disease (COVID-19) are globally at various
stages from planning to completion. Although some attempts were made to
standardize study designs, this was hindered by the ferocity of the pandemic
and the need to set up trials quickly. We take the view that a successful
treatment of COVID-19 patients (i) increases the probability of a recovery or
improvement within a certain time interval, say 28 days; (ii) aims to expedite
favourable events within this time frame; and (iii) does not increase mortality
over this time period. On this background we discuss the choice of endpoint and
its analysis. Furthermore, we consider consequences of this choice for other
design aspects including sample size and power and provide some guidance on the
application of adaptive designs in this particular context.
| stat |
Decline of war or end of positive check? Analysis of change in war size
distribution between 1816-2007 | This study examines whether there has been a decline in the risk of death by
battle during wars, testing the 'long peace' hypothesis. The analysis relies on
the Expanded War Dataset (Gleditsch, 2004) covering intra- and inter-state wars
between 1816-2007. Using untransformed data on war sizes, the estimates do not
provide empirical evidence for a decline in war over time. However, normalising
the data for global human population does illustrate a likely decline in war
from 1947 onward. The results indicate that despite strong population growth
wars have not become more severe.
| stat |
Theoretical Guarantees for Model Auditing with Finite Adversaries | Privacy concerns have led to the development of privacy-preserving approaches
for learning models from sensitive data. Yet, in practice, even models learned
with privacy guarantees can inadvertently memorize unique training examples or
leak sensitive features. To identify such privacy violations, existing model
auditing techniques use finite adversaries defined as machine learning models
with (a) access to some finite side information (e.g., a small auditing
dataset), and (b) finite capacity (e.g., a fixed neural network architecture).
Our work investigates the requirements under which an unsuccessful attempt to
identify privacy violations by a finite adversary implies that no stronger
adversary can succeed at such a task. We do so via parameters that quantify the
capabilities of the finite adversary, including the size of the neural network
employed by such an adversary and the amount of side information it has access
to as well as the regularity of the (perhaps privacy-guaranteeing) audited
model.
| stat |
The EAS approach for graphical selection consistency in vector
autoregression models | As evidenced by various recent and significant papers within the frequentist
literature, along with numerous applications in macroeconomics, genomics, and
neuroscience, there continues to be substantial interest to understand the
theoretical estimation properties of high-dimensional vector autoregression
(VAR) models. To date, however, while Bayesian VAR (BVAR) models have been
developed and studied empirically (primarily in the econometrics literature)
there exist very few theoretical investigations of the repeated sampling
properties for BVAR models in the literature. In this direction, we construct
methodology via the $\varepsilon$-$admissible$ subsets (EAS) approach for
posterior-like inference based on a generalized fiducial distribution of
relative model probabilities over all sets of active/inactive components
(graphs) of the VAR transition matrix. We provide a mathematical proof of
$pairwise$ and $strong$ graphical selection consistency for the EAS approach
for stable VAR(1) models which is robust to model misspecification, and
demonstrate numerically that it is an effective strategy in high-dimensional
settings.
| stat |
Goodness of Causal Fit | We propose a Goodness of Causal Fit (GCF) measure which depends on Pearl "do"
interventions. This is different from a measure of Goodness of Fit (GF), which
does not use interventions. Given a DAG set ${\cal G}$, to find a good $G\in
{\cal G}$, we propose plotting $GCF(G)$ versus $GF(G)$ for all $G\in {\cal G}$,
and finding a graph $G\in {\cal G}$ with a large amount of both types of
goodness.
| stat |
Variable fusion for Bayesian linear regression via spike-and-slab priors | In linear regression models, fusion of coefficients is used to identify
predictors having similar relationships with a response. This is called
variable fusion. This paper presents a novel variable fusion method in terms of
Bayesian linear regression models. We focus on hierarchical Bayesian models
based on a spike-and-slab prior approach. A spike-and-slab prior is tailored to
perform variable fusion. To obtain estimates of the parameters, we develop a
Gibbs sampler for the parameters. Simulation studies and a real data analysis
show that our proposed method achieves better performance than previous
methods.
| stat |
Exponential Dispersion Models for Overdispersed Zero-Inflated Count Data | We consider three new classes of exponential dispersion models of discrete
probability distributions which are defined by specifying their variance
functions in their mean value parameterization. In a previous paper (Bar-Lev
and Ridder, 2020a), we have developed the framework of these classes and proved
that they have some desirable properties. Each of these classes was shown to be
overdispersed and zero inflated in ascending order, making them as competitive
statistical models for those in use in statistical modeling. In this paper we
elaborate on the computational aspects of their probability mass functions.
Furthermore, we apply these classes for fitting real data sets having
overdispersed and zero-inflated statistics. Classic models based on Poisson or
negative binomial distributions show poor fits, and therefore many alternatives
have already proposed in recent years. We execute an extensive comparison with
these other proposals, from which we may conclude that our framework is a
flexible tool that gives excellent results in all cases. Moreover, in most
cases our model gives the best fit.
| stat |
Two-block vs. Multi-block ADMM: An empirical evaluation of convergence | Alternating Direction Method of Multipliers (ADMM) has become a widely used
optimization method for convex problems, particularly in the context of data
mining in which large optimization problems are often encountered. ADMM has
several desirable properties, including the ability to decompose large problems
into smaller tractable sub-problems and ease of parallelization, that are
essential in these scenarios. The most common form of ADMM is the two-block, in
which two sets of primal variables are updated alternatingly. Recent years have
seen advances in multi-block ADMM, which update more than two blocks of primal
variables sequentially. In this paper, we study the empirical question: {\em Is
two-block ADMM always comparable with sequential multi-block ADMM solving an
equivalent problem?} In the context of optimization problems arising in
multi-task learning, through a comprehensive set of experiments we surprisingly
show that multi-block ADMM consistently outperformed two-block ADMM on
optimization performance, and as a consequence on prediction performance,
across all datasets and for the entire range of dual step sizes. Our results
have an important practical implication: rather than simply using the popular
two-block ADMM, one may considerably benefit from experimenting with
multi-block ADMM applied to an equivalent problem.
| stat |
Monte Carlo simulation on the Stiefel manifold via polar expansion | Motivated by applications to Bayesian inference for statistical models with
orthogonal matrix parameters, we present $\textit{polar expansion},$ a general
approach to Monte Carlo simulation from probability distributions on the
Stiefel manifold. To bypass many of the well-established challenges of
simulating from the distribution of a random orthogonal matrix
$\boldsymbol{Q},$ we construct a distribution for an unconstrained random
matrix $\boldsymbol{X}$ such that $\boldsymbol{Q}_X,$ the orthogonal component
of the polar decomposition of $\boldsymbol{X},$ is equal in distribution to
$\boldsymbol{Q}.$ The distribution of $\boldsymbol{X}$ is amenable to Markov
chain Monte Carlo (MCMC) simulation using standard methods, and an
approximation to the distribution of $\boldsymbol{Q}$ can be recovered from a
Markov chain on the unconstrained space. When combined with modern MCMC
software, polar expansion allows for routine and flexible posterior inference
in models with orthogonal matrix parameters. We find that polar expansion with
adaptive Hamiltonian Monte Carlo is an order of magnitude more efficient than
competing MCMC approaches in a benchmark protein interaction network
application. We also propose a new approach to Bayesian functional principal
components analysis which we illustrate in a meteorological time series
application.
| stat |
Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient
Descent | A longstanding goal in deep learning research has been to precisely
characterize training and generalization. However, the often complex loss
landscapes of neural networks have made a theory of learning dynamics elusive.
In this work, we show that for wide neural networks the learning dynamics
simplify considerably and that, in the infinite width limit, they are governed
by a linear model obtained from the first-order Taylor expansion of the network
around its initial parameters. Furthermore, mirroring the correspondence
between wide Bayesian neural networks and Gaussian processes, gradient-based
training of wide neural networks with a squared loss produces test set
predictions drawn from a Gaussian process with a particular compositional
kernel. While these theoretical results are only exact in the infinite width
limit, we nevertheless find excellent empirical agreement between the
predictions of the original network and those of the linearized version even
for finite practically-sized networks. This agreement is robust across
different architectures, optimization methods, and loss functions.
| stat |
Intraday Retail Sales Forecast: An Efficient Algorithm for Quantile
Additive Modeling | With the ever increasing prominence of data in retail operations, sales
forecasting has become an essential pillar in the efficient management of
inventories. When facing high demand, the use of backroom storage and intraday
shelf replenishment is necessary to avoid stock-out. In that context, the
mandatory input for any successful replenishment policy to be implemented is
access to reliable forecasts for the sales at an intraday granularity. To that
end, we use quantile regression to adapt different patterns from one product to
the other, and we develop a stable and efficient quantile additive model
algorithm to compute sales forecasts in an intradaily context. Our algorithm is
computationally fast and is therefore suitable for use in real-time dynamic
shelf replenishment. As an illustration, we examine the case of a highly
frequented store, where the demand for various alimentary products is
accurately estimated over the day with the help of the proposed algorithm.
| stat |
Anchor regression: heterogeneous data meets causality | We consider the problem of predicting a response variable from a set of
covariates on a data set that differs in distribution from the training data.
Causal parameters are optimal in terms of predictive accuracy if in the new
distribution either many variables are affected by interventions or only some
variables are affected, but the perturbations are strong. If the training and
test distributions differ by a shift, causal parameters might be too
conservative to perform well on the above task. This motivates anchor
regression, a method that makes use of exogeneous variables to solve a
relaxation of the causal minimax problem by considering a modification of the
least-squares loss. The procedure naturally provides an interpolation between
the solutions of ordinary least squares and two-stage least squares. We prove
that the estimator satisfies predictive guarantees in terms of distributional
robustness against shifts in a linear class; these guarantees are valid even if
the instrumental variables assumptions are violated. If anchor regression and
least squares provide the same answer (anchor stability), we establish that OLS
parameters are invariant under certain distributional changes. Anchor
regression is shown empirically to improve replicability and protect against
distributional shifts.
| stat |
Clarifying causal mediation analysis for the applied researcher:
Defining effects based on what we want to learn | The incorporation of causal inference in mediation analysis has led to
theoretical and methodological advancements -- effect definitions with causal
interpretation, clarification of assumptions required for effect
identification, and an expanding array of options for effect estimation.
However, the literature on these results is fast-growing and complex, which may
be confusing to researchers unfamiliar with causal inference or unfamiliar with
mediation. The goal of this paper is to help ease the understanding and
adoption of causal mediation analysis. It starts by highlighting a key
difference between the causal inference and traditional approaches to mediation
analysis and making a case for the need for explicit causal thinking and the
causal inference approach in mediation analysis. It then explains in
as-plain-as-possible language existing effect types, paying special attention
to motivating these effects with different types of research questions, and
using concrete examples for illustration. This presentation differentiates two
perspectives (or purposes of analysis): the explanatory perspective (aiming to
explain the total effect) and the interventional perspective (asking questions
about hypothetical interventions on the exposure and mediator, or
hypothetically modified exposures). For the latter perspective, the paper
proposes tapping into a general class of interventional effects that contains
as special cases most of the usual effect types -- interventional direct and
indirect effects, controlled direct effects and also a generalized
interventional direct effect type, as well as the total effect and overall
effect. This general class allows flexible effect definitions which better
match many research questions than the standard interventional direct and
indirect effects.
| stat |
Conditional Path Analysis in Singly-Connected Path Diagrams | We extend the classical path analysis by showing that, for a singly-connected
path diagram, the partial covariance of two random variables factorizes over
the nodes and edges in the path between the variables. This result allows us to
show that Simpson's paradox cannot occur in singly-connected path diagrams.
| stat |
Receiver operating characteristic (ROC) movies, universal ROC (UROC)
curves, and coefficient of predictive ability (CPA) | Throughout science and technology, receiver operating characteristic (ROC)
curves and associated area under the curve (AUC) measures constitute powerful
tools for assessing the predictive abilities of features, markers and tests in
binary classification problems. Despite its immense popularity, ROC analysis
has been subject to a fundamental restriction, in that it applies to
dichotomous (yes or no) outcomes only. Here we introduce ROC movies and
universal ROC (UROC) curves that apply to just any ordinal or real-valued
outcome, along with a new, asymmetric coefficient of predictive ability (CPA)
measure. CPA equals the area under the UROC curve, and admits appealing
interpretations in terms of probabilities and rank based covariances. For
binary outcomes CPA equals AUC, and for pairwise distinct outcomes CPA relates
linearly to Spearman's rank correlation coefficient. ROC movies, UROC curves,
and CPA nest and generalize the tools of classical ROC analysis, and are bound
to supersede them in a wealth of applications. Their usage is illustrated in
data examples from biomedicine and meteorology, where CPA yields new insights
in the WeatherBench comparison of the predictive performance of convolutional
neural networks and physical-numerical models for weather prediction.
| stat |
Learning deep kernels for exponential family densities | The kernel exponential family is a rich class of distributions, which can be
fit efficiently and with statistical guarantees by score matching. Being
required to choose a priori a simple kernel such as the Gaussian, however,
limits its practical applicability. We provide a scheme for learning a kernel
parameterized by a deep network, which can find complex location-dependent
local features of the data geometry. This gives a very rich class of density
models, capable of fitting complex structures on moderate-dimensional problems.
Compared to deep density models fit via maximum likelihood, our approach
provides a complementary set of strengths and tradeoffs: in empirical studies,
the former can yield higher likelihoods, whereas the latter gives better
estimates of the gradient of the log density, the score, which describes the
distribution's shape.
| stat |
Taylor Moment Expansion for Continuous-Discrete Gaussian Filtering and
Smoothing | The paper is concerned with non-linear Gaussian filtering and smoothing in
continuous-discrete state-space models, where the dynamic model is formulated
as an It\^{o} stochastic differential equation (SDE), and the measurements are
obtained at discrete time instants. We propose novel Taylor moment expansion
(TME) Gaussian filter and smoother which approximate the moments of the SDE
with a temporal Taylor expansion. Differently from classical linearisation or
It\^{o}--Taylor approaches, the Taylor expansion is formed for the moment
functions directly and in time variable, not by using a Taylor expansion on the
non-linear functions in the model. We analyse the theoretical properties,
including the positive definiteness of the covariance estimate and stability of
the TME Gaussian filter and smoother. By numerical experiments, we demonstrate
that the proposed TME Gaussian filter and smoother significantly outperform the
state-of-the-art methods in terms of estimation accuracy and numerical
stability.
| stat |
Compressive Sensing Using Iterative Hard Thresholding with Low Precision
Data Representation: Theory and Applications | Modern scientific instruments produce vast amounts of data, which can
overwhelm the processing ability of computer systems. Lossy compression of data
is an intriguing solution, but comes with its own drawbacks, such as potential
signal loss, and the need for careful optimization of the compression ratio. In
this work, we focus on a setting where this problem is especially acute:
compressive sensing frameworks for interferometry and medical imaging. We ask
the following question: can the precision of the data representation be lowered
for all inputs, with recovery guarantees and practical performance? Our first
contribution is a theoretical analysis of the normalized Iterative Hard
Thresholding (IHT) algorithm when all input data, meaning both the measurement
matrix and the observation vector are quantized aggressively. We present a
variant of low precision normalized {IHT} that, under mild conditions, can
still provide recovery guarantees. The second contribution is the application
of our quantization framework to radio astronomy and magnetic resonance
imaging. We show that lowering the precision of the data can significantly
accelerate image recovery. We evaluate our approach on telescope data and
samples of brain images using CPU and FPGA implementations achieving up to a 9x
speed-up with negligible loss of recovery quality.
| stat |
Advances in Statistical Modeling of Spatial Extremes | The classical modeling of spatial extremes relies on asymptotic models (i.e.,
max-stable processes or $r$-Pareto processes) for block maxima or peaks over
high thresholds, respectively. However, at finite levels, empirical evidence
often suggests that such asymptotic models are too rigidly constrained, and
that they do not adequately capture the frequent situation where more severe
events tend to be spatially more localized. In other words, these asymptotic
models have a strong tail dependence that persists at increasingly high levels,
while data usually suggest that it should weaken instead. Another well-known
limitation of classical spatial extremes models is that they are either
computationally prohibitive to fit in high dimensions, or they need to be
fitted using less efficient techniques. In this review paper, we describe
recent progress in the modeling and inference for spatial extremes, focusing on
new models that have more flexible tail structures that can bridge asymptotic
dependence classes, and that are more easily amenable to likelihood-based
inference for large datasets. In particular, we discuss various types of random
scale constructions, as well as the conditional spatial extremes model, which
have recently been getting increasing attention within the statistics of
extremes community. We illustrate some of these new spatial models on two
different environmental applications.
| stat |
More Data Can Hurt for Linear Regression: Sample-wise Double Descent | In this expository note we describe a surprising phenomenon in
overparameterized linear regression, where the dimension exceeds the number of
samples: there is a regime where the test risk of the estimator found by
gradient descent increases with additional samples. In other words, more data
actually hurts the estimator. This behavior is implicit in a recent line of
theoretical works analyzing "double-descent" phenomenon in linear models. In
this note, we isolate and understand this behavior in an extremely simple
setting: linear regression with isotropic Gaussian covariates. In particular,
this occurs due to an unconventional type of bias-variance tradeoff in the
overparameterized regime: the bias decreases with more samples, but variance
increases.
| stat |
Bayesian semiparametric modelling of phase-varying point processes | We propose a Bayesian semiparametric approach for registration of multiple
point processes. Our approach entails modelling the mean measures of the
phase-varying point processes with a Bernstein-Dirichlet prior, which induces a
prior on the space of all warp functions. Theoretical results on the support of
the induced priors are derived, and posterior consistency is obtained under
mild conditions. Numerical experiments suggest a good performance of the
proposed methods, and a climatology real-data example is used to showcase how
the method can be employed in practice.
| stat |
Bayesian Experimental Design for Finding Reliable Level Set under Input
Uncertainty | In the manufacturing industry, it is often necessary to repeat expensive
operational testing of machine in order to identify the range of input
conditions under which the machine operates properly. Since it is often
difficult to accurately control the input conditions during the actual usage of
the machine, there is a need to guarantee the performance of the machine after
properly incorporating the possible variation in input conditions. In this
paper, we formulate this practical manufacturing scenario as an Input Uncertain
Reliable Level Set Estimation (IU-rLSE) problem, and provide an efficient
algorithm for solving it. The goal of IU-rLSE is to identify the input range in
which the outputs smaller/greater than a desired threshold can be obtained with
high probability when the input uncertainty is properly taken into
consideration. We propose an active learning method to solve the IU-rLSE
problem efficiently, theoretically analyze its accuracy and convergence, and
illustrate its empirical performance through numerical experiments on
artificial and real data.
| stat |
A Case Study of Promoting Informal Inferential Reasoning in Learning
Sampling Distribution for High School Students | Drawing inference from data is an important skill for students to understand
their everyday life, so that the sampling distribution as a central topic in
statistical inference is necessary to be learned by the students. However,
little is known about how to teach the topic for high school students,
especially in Indonesian context. Therefore, the present study provides a
teaching experiment to support the students' informal inferential reasoning in
understanding the sampling distribution, as well as the students' perceptions
toward the teaching experiment. The subjects in the present study were three
11th-grader of one private school in Yogyakarta majoring in mathematics and
natural science. The method of data collection was direct observation of
sampling distribution learning process, interviews, and documentation. The
present study found that that informal inferential reasoning with problem-based
learning using contextual problems and real data could support the students to
understand the sampling distribution, and they also gave positive responses
about their learning experience.
| stat |
Boosting Local Causal Discovery in High-Dimensional Expression Data | We study the performance of Local Causal Discovery (LCD), a simple and
efficient constraint-based method for causal discovery, in predicting causal
effects in large-scale gene expression data. We construct practical estimators
specific to the high-dimensional regime. Inspired by the ICP algorithm, we use
an optional preselection method and two different statistical tests.
Empirically, the resulting LCD estimator is seen to closely approach the
accuracy of ICP, the state-of-the-art method, while it is algorithmically
simpler and computationally more efficient.
| stat |
The R Package stagedtrees for Structural Learning of Stratified Staged
Trees | stagedtrees is an R package which includes several algorithms for learning
the structure of staged trees and chain event graphs from data. Score-based and
clustering-based algorithms are implemented, as well as various functionalities
to plot the models and perform inference. The capabilities of stagedtrees are
illustrated using mainly two datasets both included in the package or bundled
in R.
| stat |
Docs are ROCs: A simple off-the-shelf approach for estimating average
human performance in diagnostic studies | Estimating average human performance has been performed inconsistently in
research in diagnostic medicine. This has been particularly apparent in the
field of medical artificial intelligence, where humans are often compared
against AI models in multi-reader multi-case studies, and commonly reported
metrics such as the pooled or average human sensitivity and specificity will
systematically underestimate the performance of human experts. We present the
use of summary receiver operating characteristic curve analysis, a technique
commonly used in the meta-analysis of diagnostic test accuracy studies, as a
sensible and methodologically robust alternative. We describe the motivation
for using these methods and present results where we apply these meta-analytic
techniques to a handful of prominent medical AI studies.
| stat |
Effect heterogeneity and variable selection for standardizing causal
effects to a target population | The participants in randomized trials and other studies used for causal
inference are often not representative of the populations seen by clinical
decision-makers. To account for differences between populations, researchers
may consider standardizing results to a target population. We discuss several
different types of homogeneity conditions that are relevant for
standardization: Homogeneity of effect measures, homogeneity of counterfactual
outcome state transition parameters, and homogeneity of counterfactual
distributions. Each of these conditions can be used to show that a particular
standardization procedure will result in unbiased estimates of the effect in
the target population, given assumptions about the relevant scientific context.
We compare and contrast the homogeneity conditions, in particular their
implications for selection of covariates for standardization and their
implications for how to compute the standardized causal effect in the target
population. While some of the recently developed counterfactual approaches to
generalizability rely upon homogeneity conditions that avoid many of the
problems associated with traditional approaches, they often require adjustment
for a large (and possibly unfeasible) set of covariates.
| stat |
Tables of Quantiles of the Distribution of the Empirical Chiral Index in
the Case of the Uniform Law and in the Case of the Normal Law | The empirical distribution of the chiral index is simulated for various
sample sizes for the uniform law and and for the normal law. The estimated
quantiles $K_{0.90}$, $K_{0.95}$, $K_{0.98}$, and $K_{0.99}$, are tabulated for
use in symmetry testing in the uniform case and in the normal case.
| stat |
Assessing Bayes factor surfaces using interactive visualization and
computer surrogate modeling | Bayesian model selection provides a natural alternative to classical
hypothesis testing based on p-values. While many papers mention that Bayesian
model selection is frequently sensitive to prior specification on the
parameters, there are few practical strategies to assess and report this
sensitivity. This article has two goals. First, we aim educate the broader
statistical community about the extent of potential sensitivity through
visualization of the Bayes factor surface. The Bayes factor surface shows the
value a Bayes factor takes (usually on the log scale) as a function of
user-specified hyperparameters. We provide interactive visualization through an
R shiny application that allows the user to explore sensitivity in Bayes factor
over a range of hyperparameter settings in a familiar regression setting. We
compare the surface with three automatic procedures. Second, we suggest
surrogate modeling via Gaussian processes (GPs) to visualize the Bayes factor
surface in situations where computation of Bayes factors is expensive. That is,
we treat Bayes factor calculation as a computer simulation experiment. In this
context, we provide a fully reproducible example using accessible GP libraries
to augment an important study of the influence of outliers in empirical
finance. We suggest Bayes factor surfaces are valuable for scientific reporting
since they (i) increase transparency, making potential instability in Bayes
factors easy to visualize, (ii) generalize to simple and more complicated
examples, and (iii) provide a path for researchers to assess the impact of
prior choice on modeling decisions in a wide variety research areas.
| stat |
The global optimum of shallow neural network is attained by ridgelet
transform | We prove that the global minimum of the backpropagation (BP) training problem
of neural networks with an arbitrary nonlinear activation is given by the
ridgelet transform. A series of computational experiments show that there
exists an interesting similarity between the scatter plot of hidden parameters
in a shallow neural network after the BP training and the spectrum of the
ridgelet transform. By introducing a continuous model of neural networks, we
reduce the training problem to a convex optimization in an infinite dimensional
Hilbert space, and obtain the explicit expression of the global optimizer via
the ridgelet transform.
| stat |
Bayesian waveform-based calibration of high-pressure acoustic emission
systems with ball drop measurements | Acoustic emission (AE) is a widely used technology to study source mechanisms
and material properties during high-pressure rock failure experiments. It is
important to understand the physical quantities that acoustic emission sensors
measure, as well as the response of these sensors as a function of frequency.
This study calibrates the newly built AE system in the MIT Rock Physics
Laboratory using a ball-bouncing system. Full waveforms of multi-bounce events
due to ball drops are used to infer the transfer function of lead zirconate
titanate (PZT) sensors in high pressure environments. Uncertainty in the sensor
transfer functions is quantified using a waveform-based Bayesian approach. The
quantification of \textit{in situ} sensor transfer functions makes it possible
to apply full waveform analysis for acoustic emissions at high pressures.
| stat |
Towards a Theoretical Understanding of the Robustness of Variational
Autoencoders | We make inroads into understanding the robustness of Variational Autoencoders
(VAEs) to adversarial attacks and other input perturbations. While previous
work has developed algorithmic approaches to attacking and defending VAEs,
there remains a lack of formalization for what it means for a VAE to be robust.
To address this, we develop a novel criterion for robustness in probabilistic
models: $r$-robustness. We then use this to construct the first theoretical
results for the robustness of VAEs, deriving margins in the input space for
which we can provide guarantees about the resulting reconstruction. Informally,
we are able to define a region within which any perturbation will produce a
reconstruction that is similar to the original reconstruction. To support our
analysis, we show that VAEs trained using disentangling methods not only score
well under our robustness metrics, but that the reasons for this can be
interpreted through our theoretical results.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.