title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Predicting play calls in the National Football League using hidden
Markov models | In recent years, data-driven approaches have become a popular tool in a
variety of sports to gain an advantage by, e.g., analysing potential strategies
of opponents. Whereas the availability of play-by-play or player tracking data
in sports such as basketball and baseball has led to an increase of sports
analytics studies, equivalent datasets for the National Football League (NFL)
were not freely available for a long time. In this contribution, we consider a
comprehensive play-by-play NFL dataset provided by www.kaggle.com, comprising
289,191 observations in total, to predict play calls in the NFL using hidden
Markov models. The resulting out-of-sample prediction accuracy for the 2018 NFL
season is 71.5%, which is substantially higher compared to similar studies on
play call predictions in the NFL.
| stat |
Improving the Power of the Randomization Test | We consider the problem of evaluating designs for a two-arm randomized
experiment with the criterion being the power of the randomization test for the
one-sided null hypothesis. Our evaluation assumes a response that is linear in
one observed covariate, an unobserved component and an additive treatment
effect where the only randomness comes from the treatment allocations. It is
well-known that the power depends on the allocations' imbalance in the observed
covariate and this is the reason for the classic restricted designs such as
rerandomization. We show that power is also affected by two other design
choices: the number of allocations in the design and the degree of linear
dependence among the allocations. We prove that the more allocations, the
higher the power and the lower the variability in the power. Designs that
feature greater independence of allocations are also shown to have higher
performance.
Our theoretical findings and extensive simulation studies imply that the
designs with the highest power provide thousands of highly independent
allocations that each provide nominal imbalance in the observed covariates.
These high powered designs exhibit less randomization than complete
randomization and more randomization than recently proposed designs based on
numerical optimization. Model choices for a practicing experimenter are
rerandomization and greedy pair switching, where both outperform complete
randomization and numerical optimization. The tradeoff we find also provides a
means to specify the imbalance threshold parameter when rerandomizing.
| stat |
Minimax designs for causal effects in temporal experiments with
treatment habituation | Randomized experiments are the gold standard for estimating the causal
effects of an intervention. In the simplest setting, each experimental unit is
randomly assigned to receive treatment or control, and then the outcomes in
each treatment arm are compared. In many settings, however, randomized
experiments need to be executed over several time periods such that treatment
assignment happens at each time period. In such temporal experiments, it has
been observed that the effects of an intervention on a given unit may be large
when the unit is first exposed to it, but then it often attenuates, or even
vanishes, after repeated exposures. This phenomenon is typically due to units'
habituation to the intervention, or some other general form of learning, such
as when users gradually start to ignore repeated mails sent by a promotional
campaign. This paper proposes randomized designs for estimating causal effects
in temporal experiments when habituation is present. We show that our designs
are minimax optimal in a large class of practical designs. Our analysis is
based on the randomization framework of causal inference, and imposes no
parametric modeling assumptions on the outcomes.
| stat |
The computational asymptotics of Gaussian variational inference | Variational inference is a popular alternative to Markov chain Monte Carlo
methods that constructs a Bayesian posterior approximation by minimizing a
discrepancy to the true posterior within a pre-specified family. This converts
Bayesian inference into an optimization problem, enabling the use of simple and
scalable stochastic optimization algorithms. However, a key limitation of
variational inference is that the optimal approximation is typically not
tractable to compute; even in simple settings the problem is nonconvex. Thus,
recently developed statistical guarantees -- which all involve the (data)
asymptotic properties of the optimal variational distribution -- are not
reliably obtained in practice. In this work, we provide two major
contributions: a theoretical analysis of the asymptotic convexity properties of
variational inference in the popular setting with a Gaussian family; and
consistent stochastic variational inference (CSVI), an algorithm that exploits
these properties to find the optimal approximation in the asymptotic regime.
CSVI consists of a tractable initialization procedure that finds the local
basin of the optimal solution, and a scaled gradient descent algorithm that
stays locally confined to that basin. Experiments on nonconvex synthetic and
real-data examples show that compared with standard stochastic gradient
descent, CSVI improves the likelihood of obtaining the globally optimal
posterior approximation.
| stat |
Excess deaths, baselines, Z-scores, P-scores and peaks | The recent Covid-19 epidemic has lead to comparisons of the countries
suffering from it. These are based on the number of excess deaths attributed
either directly or indirectly to the epidemic. Unfortunately the data on which
such comparisons rely are often incomplete and unreliable. This article
discusses problems of interpretation of data even when the data is largely
accurate and delayed by at most two to three weeks. This applies to the Office
of National Statistics in the UK, the Statistisches Bundesamt in Germany and
the Belgian statistical office Statbel. The data in the article is taken from
these three sources. The number of excess deaths is defined as the number of
deaths minus the baseline, the definition of which varies from country to
country. In the UK it is the average number of deaths over the last five years,
in Germany it is over the last four years and in Belgium over the last 11
years. This means that in all cases the individual baselines depend strongly on
the timing and intensity of adverse factors such as past influenza epidemics
and heat waves. This makes cross-country comparisons difficult. A baseline
defined as the number the number of deaths in the absence of adverse factors
can be operationalized by taking say the 10\% quantile of the number of deaths.
This varies little over time and European countries within given age groups. It
therefore enables more robust and accurate comparisons of different countries.
The article criticizes the use of Z-scores which distort the comparison between
countries. Finally the problem of describing past epidemics by their timing,
that is start and finish and time of the maximum, and by their effect, the
height of the maximum and the total number of deaths, is considered.
| stat |
Random projections: data perturbation for classification problems | Random projections offer an appealing and flexible approach to a wide range
of large-scale statistical problems. They are particularly useful in
high-dimensional settings, where we have many covariates recorded for each
observation. In classification problems there are two general techniques using
random projections. The first involves many projections in an ensemble -- the
idea here is to aggregate the results after applying different random
projections, with the aim of achieving superior statistical accuracy. The
second class of methods include hashing and sketching techniques, which are
straightforward ways to reduce the complexity of a problem, perhaps therefore
with a huge computational saving, while approximately preserving the
statistical efficiency.
| stat |
Bayesian Analysis of Static Light Scattering Data for Globular Proteins | Static light scattering is a popular physical chemistry technique that
enables calculation of physical attributes such as the radius of gyration and
the second virial coefficient for a macromolecule (e.g., a polymer or a
protein) in solution. The second virial coefficient is a physical quantity that
characterizes the magnitude and sign of pairwise interactions between
particles, and hence is related to aggregation propensity, a property of
considerable scientific and practical interest. Estimating the second virial
coefficient from experimental data is challenging due both to the degree of
precision required and the complexity of the error structure involved. In
contrast to conventional approaches based on heuristic OLS estimates, Bayesian
inference for the second virial coefficient allows explicit modeling of error
processes, incorporation of prior information, and the ability to directly test
competing physical models. Here, we introduce a fully Bayesian model for static
light scattering experiments on small-particle systems, with joint inference
for concentration, index of refraction, oligomer size, and the second virial
coefficient. We apply our proposed model to study the aggregation behavior of
hen egg-white lysozyme and human gammaS-crystallin using in-house experimental
data. Based on these observations, we also perform a simulation study on the
primary drivers of uncertainty in this family of experiments, showing in
particular the potential for improved monitoring and control of concentration
to aid inference.
| stat |
Graph Convolution with Low-rank Learnable Local Filters | Geometric variations like rotation, scaling, and viewpoint changes pose a
significant challenge to visual understanding. One common solution is to
directly model certain intrinsic structures, e.g., using landmarks. However, it
then becomes non-trivial to build effective deep models, especially when the
underlying non-Euclidean grid is irregular and coarse. Recent deep models using
graph convolutions provide an appropriate framework to handle such
non-Euclidean data, but many of them, particularly those based on global graph
Laplacians, lack expressiveness to capture local features required for
representation of signals lying on the non-Euclidean grid. The current paper
introduces a new type of graph convolution with learnable low-rank local
filters, which is provably more expressive than previous spectral graph
convolution methods. The model also provides a unified framework for both
spectral and spatial graph convolutions. To improve model robustness,
regularization by local graph Laplacians is introduced. The representation
stability against input graph data perturbation is theoretically proved, making
use of the graph filter locality and the local graph regularization.
Experiments on spherical mesh data, real-world facial expression
recognition/skeleton-based action recognition data, and data with simulated
graph noise show the empirical advantage of the proposed model.
| stat |
Manifold Density Estimation via Generalized Dequantization | Density estimation is an important technique for characterizing distributions
given observations. Much existing research on density estimation has focused on
cases wherein the data lies in a Euclidean space. However, some kinds of data
are not well-modeled by supposing that their underlying geometry is Euclidean.
Instead, it can be useful to model such data as lying on a {\it manifold} with
some known structure. For instance, some kinds of data may be known to lie on
the surface of a sphere. We study the problem of estimating densities on
manifolds. We propose a method, inspired by the literature on "dequantization,"
which we interpret through the lens of a coordinate transformation of an
ambient Euclidean space and a smooth manifold of interest. Using methods from
normalizing flows, we apply this method to the dequantization of smooth
manifold structures in order to model densities on the sphere, tori, and the
orthogonal group.
| stat |
Statistical Properties of Sanitized Results from Differentially Private
Laplace Mechanism with Univariate Bounding Constraints | Protection of individual privacy is a common concern when releasing and
sharing data and information. Differential privacy (DP) formalizes privacy in
probabilistic terms without making assumptions about the background knowledge
of data intruders, and thus provides a robust concept for privacy protection.
Practical applications of DP involve development of differentially private
mechanisms to generate sanitized results at a pre-specified privacy budget. For
the sanitization of statistics with publicly known bounds such as proportions
and correlation coefficients, the bounding constraints will need to be
incorporated in the differentially private mechanisms. There has been little
work on examining the consequences of the bounding constraints on the accuracy
of sanitized results and the statistical inferences of the population
parameters based on the sanitized results. In this paper, we formalize the
differentially private truncated and boundary inflated truncated (BIT)
procedures for releasing statistics with publicly known bounding constraints.
The impacts of the truncated and BIT Laplace procedures on the statistical
accuracy and validity of sanitized statistics are evaluated both theoretically
and empirically via simulation studies.
| stat |
Regularised Least-Squares Regression with Infinite-Dimensional Output
Space | We present some learning theory results on vector-valued reproducing kernel
Hilbert space (RKHS) regression, where the input space is allowed to be
non-compact and the output space is a (possibly infinite-dimensional) Hilbert
space.
| stat |
Adaptive Generalized Logit-Normal Distributions for Wind Power
Short-Term Forecasting | There is increasing interest in very short-term and higher-resolution wind
power forecasting (from minutes to hours ahead), especially offshore.
Statistical methods are of utmost relevance, since weather forecasts cannot be
informative for those lead times. Those approaches ought to account for the
fact that wind power generation as a stochastic process is nonstationary,
double-bounded (by zero and the nominal power of the turbine) and non-linear.
Accommodating those aspects may lead to improving both point and probabilistic
forecasts. We propose here to focus on generalized logit-normal distributions,
which are naturally suitable and flexible for double-bounded and non-linear
processes. Relevant parameters are estimated via maximum likelihood inference.
Both batch and online versions of the estimation approach are described -- the
online version permitting to additionally handle non-stationarity through the
variation of distribution parameters. The approach is applied and analysed on
the test case of the Anholt offshore wind farm in Denmark, with emphasis placed
on 10-min-ahead point and probabilistic forecasts.
| stat |
Soft Random Graphs in Probabilistic Metric Spaces & Inter-graph Distance | We present a new method for learning Soft Random Geometric Graphs (SRGGs),
drawn in probabilistic metric spaces, with the connection function of the graph
defined as the marginal posterior probability of an edge random variable, given
the correlation between the nodes connected by that edge. In fact, this
inter-node correlation matrix is itself a random variable in our learning
strategy, and we learn this by identifying each node as a random variable,
measurements of which comprise a column of a given multivariate dataset. We
undertake inference with Metropolis with a 2-block update scheme. The SRGG is
shown to be generated by a non-homogeneous Poisson point process, the intensity
of which is location-dependent. Given the multivariate dataset, likelihood of
the inter-column correlation matrix is attained following achievement of a
closed-form marginalisation over all inter-row correlation matrices. Distance
between a pair of graphical models learnt given the respective datasets, offers
the absolute correlation between the given datasets; such inter-graph distance
computation is our ultimate objective, and is achieved using a newly introduced
metric that resembles an uncertainty-normalised Hellinger distance between
posterior probabilities of the two learnt SRGGs, given the respective datasets.
Two sets of empirical illustrations on real data are undertaken, and
application to simulated data is included to exemplify the effect of
incorporating measurement noise in the learning of a graphical model.
| stat |
Targeted Fused Ridge Estimation of Inverse Covariance Matrices from
Multiple High-Dimensional Data Classes | We consider the problem of jointly estimating multiple inverse covariance
matrices from high-dimensional data consisting of distinct classes. An
$\ell_2$-penalized maximum likelihood approach is employed. The suggested
approach is flexible and generic, incorporating several other
$\ell_2$-penalized estimators as special cases. In addition, the approach
allows specification of target matrices through which prior knowledge may be
incorporated and which can stabilize the estimation procedure in
high-dimensional settings. The result is a targeted fused ridge estimator that
is of use when the precision matrices of the constituent classes are believed
to chiefly share the same structure while potentially differing in a number of
locations of interest. It has many applications in (multi)factorial study
designs. We focus on the graphical interpretation of precision matrices with
the proposed estimator then serving as a basis for integrative or meta-analytic
Gaussian graphical modeling. Situations are considered in which the classes are
defined by data sets and subtypes of diseases. The performance of the proposed
estimator in the graphical modeling setting is assessed through extensive
simulation experiments. Its practical usability is illustrated by the
differential network modeling of 12 large-scale gene expression data sets of
diffuse large B-cell lymphoma subtypes. The estimator and its related
procedures are incorporated into the R-package rags2ridges.
| stat |
Reducing bias and increasing utility by federated generative modeling of
medical images using a centralized adversary | We introduce FELICIA (FEderated LearnIng with a CentralIzed Adversary) a
generative mechanism enabling collaborative learning. In particular, we show
how a data owner with limited and biased data could benefit from other data
owners while keeping data from all the sources private. This is a common
scenario in medical image analysis where privacy legislation prevents data from
being shared outside local premises. FELICIA works for a large family of
Generative Adversarial Networks (GAN) architectures including vanilla and
conditional GANs as demonstrated in this work. We show that by using the
FELICIA mechanism, a data owner with limited image samples can generate
high-quality synthetic images with high utility while neither data owners has
to provide access to its data. The sharing happens solely through a central
discriminator that has access limited to synthetic data. Here, utility is
defined as classification performance on a real test set. We demonstrate these
benefits on several realistic healthcare scenarions using benchmark image
datasets (MNIST, CIFAR-10) as well as on medical images for the task of skin
lesion classification. With multiple experiments, we show that even in the
worst cases, combining FELICIA with real data gracefully achieves performance
on par with real data while most results significantly improves the utility.
| stat |
Evaluating one-shot tournament predictions | We introduce the Tournament Rank Probability Score (TRPS) as a measure to
evaluate and compare pre-tournament predictions, where predictions of the full
tournament results are required to be available before the tournament begins.
The TRPS handles partial ranking of teams, gives credit to predictions that are
only slightly wrong, and can be modified with weights to stress the importance
of particular features of the tournament prediction. Thus, the Tournament Rank
Prediction Score is more flexible than the commonly preferred log loss score
for such tasks. In addition, we show how predictions from historic tournaments
can be optimally combined into ensemble predictions in order to maximize the
TRPS for a new tournament.
| stat |
Estimation error analysis of deep learning on the regression problem on
the variable exponent Besov space | Deep learning has achieved notable success in various fields, including image
and speech recognition. One of the factors in the successful performance of
deep learning is its high feature extraction ability. In this study, we focus
on the adaptivity of deep learning; consequently, we treat the variable
exponent Besov space, which has a different smoothness depending on the input
location $x$. In other words, the difficulty of the estimation is not uniform
within the domain. We analyze the general approximation error of the variable
exponent Besov space and the approximation and estimation errors of deep
learning. We note that the improvement based on adaptivity is remarkable when
the region upon which the target function has less smoothness is small and the
dimension is large. Moreover, the superiority to linear estimators is shown
with respect to the convergence rate of the estimation error.
| stat |
How to Control the Error Rates of Binary Classifiers | The traditional binary classification framework constructs classifiers which
may have good accuracy, but whose false positive and false negative error rates
are not under users' control. In many cases, one of the errors is more severe
and only the classifiers with the corresponding rate lower than the predefined
threshold are acceptable. In this study, we combine binary classification with
statistical hypothesis testing to control the target error rate of already
trained classifiers. In particular, we show how to turn binary classifiers into
statistical tests, calculate the classification p-values, and use them to limit
the target error rate.
| stat |
Robust Lasso-Zero for sparse corruption and model selection with missing
covariates | We propose Robust Lasso-Zero, an extension of the Lasso-Zero methodology
[Descloux and Sardy, 2018], initially introduced for sparse linear models, to
the sparse corruptions problem. We give theoretical guarantees on the sign
recovery of the parameters for a slightly simplified version of the estimator,
called Thresholded Justice Pursuit. The use of Robust Lasso-Zero is showcased
for variable selection with missing values in the covariates. In addition to
not requiring the specification of a model for the covariates, nor estimating
their covariance matrix or the noise variance, the method has the great
advantage of handling missing not-at random values without specifying a
parametric model. Numerical experiments and a medical application underline the
relevance of Robust Lasso-Zero in such a context with few available
competitors. The method is easy to use and implemented in the R library lass0.
| stat |
The Dependent Dirichlet Process and Related Models | Standard regression approaches assume that some finite number of the response
distribution characteristics, such as location and scale, change as a
(parametric or nonparametric) function of predictors. However, it is not always
appropriate to assume a location/scale representation, where the error
distribution has unchanging shape over the predictor space. In fact, it often
happens in applied research that the distribution of responses under study
changes with predictors in ways that cannot be reasonably represented by a
finite dimensional functional form. This can seriously affect the answers to
the scientific questions of interest, and therefore more general approaches are
indeed needed. This gives rise to the study of fully nonparametric regression
models. We review some of the main Bayesian approaches that have been employed
to define probability models where the complete response distribution may vary
flexibly with predictors. We focus on developments based on modifications of
the Dirichlet process, historically termed dependent Dirichlet processes, and
some of the extensions that have been proposed to tackle this general problem
using nonparametric approaches.
| stat |
Effects of winter climate on high speed passenger trains in
Botnia-Atlantica region | Harsh winter climate can cause various problems for both public and private
sectors in Sweden, especially in the northern part for railway industry. To
have a better understanding of winter climate impacts, this study investigates
effects of the winter climate including atmospheric icing on the performance of
high speed passenger trains in the Botnia-Atlantica region. The investigation
is done with train operational data together with simulated weather data from
the Weather Research and Forecast model over January - February 2017.
Two different measurements of the train performance are analysed. One is
cumulative delay which measures the increment in delay in terms of running time
within two consecutive measuring spots, the other is current delay which is the
delay in terms of arrival time at each measuring spot compared to the schedule.
Cumulative delay is investigated through a Cox model and the current delay is
studied using a Markov chain model.
The results show that the weather factors have impacts on the train
performance. Therein temperature and humidity have significant impacts on both
the occurrence of cumulative delay and the transition probabilities between
(current) delayed and non-delayed states.
| stat |
Rademacher upper bounds for cross-validation errors with an application
to the lasso | We establish a general upper bound for $K$-fold cross-validation ($K$-CV)
errors that can be adapted to many $K$-CV-based estimators and learning
algorithms. Based on Rademacher complexity of the model and the
Orlicz-$\Psi_{\nu}$ norm of the error process, the CV error upper bound applies
to both light-tail and heavy-tail error distributions. We also extend the CV
error upper bound to $\beta$-mixing data using the technique of independent
blocking. We provide a Python package (\texttt{CVbound},
\url{https://github.com/isaac2math}) for computing the CV error upper bound in
$K$-CV-based algorithms. Using the lasso as an example, we demonstrate in
simulations that the upper bounds are tight and stable across different
parameter settings and random seeds. As well as accurately bounding the CV
errors for the lasso, the minimizer of the new upper bounds can be used as a
criterion for variable selection. Compared with the CV-error minimizer,
simulations show that tuning the lasso penalty parameter according to the
minimizer of the upper bound yields a more sparse and more stable model that
retains all of the relevant variables.
| stat |
Exponential moving average versus moving exponential average | In this note we discuss the mathematical tools to define trend indicators
which are used to describe market trends. We explain the relation between
averages and moving averages on the one hand and the so called exponential
moving average (EMA) on the other hand. We present a lot of examples and give
the definition of the most frequently used trend indicator, the MACD, and
discuss its properties.
| stat |
Estimating the reciprocal of a binomial proportion | As a classic parameter from the binomial distribution, the binomial
proportion has been well studied in the literature owing to its wide range of
applications. In contrast, the reciprocal of the binomial proportion, also
known as the inverse proportion, is often overlooked, even though it also plays
an important role in various fields including clinical studies and random
sampling. The maximum likelihood estimator of the inverse proportion suffers
from the zero-event problem, and to overcome it, alternative methods have been
developed in the literature. Nevertheless, there is little work addressing the
optimality of the existing estimators, as well as their practical performance
comparison. Inspired by this, we propose to further advance the literature by
developing an optimal estimator for the inverse proportion in a family of
shrinkage estimators. We further derive the explicit and approximate formulas
for the optimal shrinkage parameter under different settings. Simulation
studies show that the performance of our new estimator performs better than, or
as well as, the existing competitors in most practical settings. Finally, to
illustrate the usefulness of our new method, we also revisit a recent
meta-analysis on COVID-19 data for assessing the relative risks of physical
distancing on the infection of coronavirus, in which six out of seven studies
encounter the zero-event problem.
| stat |
Spectral Pruning: Compressing Deep Neural Networks via Spectral Analysis
and its Generalization Error | Compression techniques for deep neural network models are becoming very
important for the efficient execution of high-performance deep learning systems
on edge-computing devices. The concept of model compression is also important
for analyzing the generalization error of deep learning, known as the
compression-based error bound. However, there is still huge gap between a
practically effective compression method and its rigorous background of
statistical learning theory. To resolve this issue, we develop a new
theoretical framework for model compression and propose a new pruning method
called {\it spectral pruning} based on this framework. We define the ``degrees
of freedom'' to quantify the intrinsic dimensionality of a model by using the
eigenvalue distribution of the covariance matrix across the internal nodes and
show that the compression ability is essentially controlled by this quantity.
Moreover, we present a sharp generalization error bound of the compressed model
and characterize the bias--variance tradeoff induced by the compression
procedure. We apply our method to several datasets to justify our theoretical
analyses and show the superiority of the the proposed method.
| stat |
Analyzing MCMC Output | Markov chain Monte Carlo (MCMC) is a sampling-based method for estimating
features of probability distributions. MCMC methods produce a serially
correlated, yet representative, sample from the desired distribution. As such
it can be difficult to know when the MCMC method is producing reliable results.
We introduce some fundamental methods for ensuring a trustworthy simulation
experiment. In particular, we present a workflow for output analysis in MCMC
providing estimators, approximate sampling distributions, stopping rules, and
visualization tools.
| stat |
Pleiotropy robust methods for multivariable Mendelian randomization | Mendelian randomization is a powerful tool for inferring the presence, or
otherwise, of causal effects from observational data. However, the nature of
genetic variants is such that pleiotropy remains a barrier to valid causal
effect estimation. There are many options in the literature for pleiotropy
robust methods when studying the effects of a single risk factor on an outcome.
However, there are few pleiotropy robust methods in the multivariable setting,
that is, when there are multiple risk factors of interest. In this paper we
introduce three methods which build on common approaches in the univariable
setting: MVMR-Robust; MVMR-Median; and MVMR-Lasso. We discuss the properties of
each of these methods and examine their performance in comparison to existing
approaches in a simulation study. MVMR-Robust is shown to outperform existing
outlier robust approaches when there are low levels of pleiotropy. MVMR-Lasso
provides the best estimation in terms of mean squared error for moderate to
high levels of pleiotropy, and can provide valid inference in a three sample
setting. MVMR-Median performs well in terms of estimation across all scenarios
considered, and provides valid inference up to a moderate level of pleiotropy.
We demonstrate the methods in an applied example looking at the effects of
intelligence, education and household income on the risk of Alzheimer's
disease.
| stat |
Measurement Errors in Semiparametric Generalized Linear Models | Regression models that ignore measurement error in predictors may produce
highly biased estimates leading to erroneous inferences. It is well known that
it is extremely difficult to take measurement error into account in Gaussian
nonparametric regression. This problem becomes tremendously more difficult when
considering other families such as logistic regression, Poisson and
negative-binomial. For the first time, we present a method aiming to correct
for measurement error when estimating regression functions flexibly covering
virtually all distributions and link functions regularly considered in
generalized linear models. This approach depends on approximating the first and
the second moment of the response after integrating out the true unobserved
predictors in a semiparametric generalized linear model. Unlike previous
methods, this method is not restricted to truncated splines and can utilize
various basis functions. Through extensive simulation studies, we study the
performance of our method under many scenarios.
| stat |
High-dimensional inference using the extremal skew-$t$ process | Max-stable processes are a popular tool for the study of environmental
extremes, and the extremal skew-$t$ process is a general model that allows for
a flexible extremal dependence structure. For inference on max-stable processes
with high-dimensional data, exact likelihood-based estimation is
computationally intractable. Composite likelihoods, using lower dimensional
components, and Stephenson-Tawn likelihoods, using occurrence times of maxima,
are both attractive methods to circumvent this issue for moderate dimensions.
In this article we establish the theoretical formulae for simulations of and
inference for the extremal skew-$t$ process. We also incorporate the
Stephenson-Tawn concept into the composite likelihood framework, giving greater
statistical and computational efficiency for higher-order composite
likelihoods. We compare 2-way (pairwise), 3-way (triplewise), 4-way, 5-way and
10-way composite likelihoods for models of up to 100 dimensions. Furthermore,
we propose cdf approximations for the Stephenson-Tawn likelihood function,
leading to large computational gains, and enabling accurate fitting of models
in large dimensions in only a few minutes. We illustrate our methodology with
an application to a 90-dimensional temperature dataset from Melbourne,
Australia.
| stat |
Multi-Class Multiple Instance Learning for Predicting Precursors to
Aviation Safety Events | In recent years, there has been a rapid growth in the application of machine
learning techniques that leverage aviation data collected from commercial
airline operations to improve safety. Anomaly detection and predictive
maintenance have been the main targets for machine learning applications.
However, this paper focuses on the identification of precursors, which is a
relatively newer application. Precursors are events correlated with adverse
events that happen prior to the adverse event itself. Therefore, precursor
mining provides many benefits including understanding the reasons behind a
safety incident and the ability to identify signatures, which can be tracked
throughout a flight to alert the operators of the potential for an adverse
event in the future. This work proposes using the multiple-instance learning
(MIL) framework, a weakly supervised learning task, combined with carefully
designed binary classifier leveraging a Multi-Head Convolutional Neural
Network-Recurrent Neural Network (MHCNN-RNN) architecture. Multi-class
classifiers are then created and compared, enabling the prediction of different
adverse events for any given flight by combining binary classifiers, and by
modifying the MHCNN-RNN to handle multiple outputs. Results obtained showed
that the multiple binary classifiers perform better and are able to accurately
forecast high speed and high path angle events during the approach phase.
Multiple binary classifiers are also capable of determining the aircraft's
parameters that are correlated to these events. The identified parameters can
be considered precursors to the events and may be studied/tracked further to
prevent these events in the future.
| stat |
Two-stage approach for the inference of the source of high-dimension and
complex chemical data in forensic science | Forensic scientists are often criticised for the lack of quantitative support
for the conclusions of their examinations. While scholars advocate for the use
of a Bayes factor to quantify the weight of forensic evidence, it is often
impossible to assign the necessary probability measures to perform
likelihood-based inference for high-dimensional and complex data. To address
this issue, we revisit a two-stage inference framework and leverage the
properties of kernel functions to offer a method that allows for statistically
supporting the inference of the identity of source of sets of trace and control
objects by way of a single test. Our method is generic in that it can be easily
tailored to any type of data encountered in forensic science or pattern
recognition, and our method does not depend on the dimension or the type of the
considered data. The application of our method to paint evidence shows that
this type of evidence carries substantial probative value. Finally, our
approach can easily be extended to other evidence types such as glass, fibres
and dust.
| stat |
A Dynamic Network for Cryptocurrencies | Cryptocurrencies return cross-predictability yields information on risk
propagation and market segmentation. To explore these effects, we build a
dynamic network of cryptocurrencies based on the evolution of return
cross-predictability and develop a dynamic covariate-assisted spectral
clustering method to consistently estimate the latent group membership of
cryptocurrencies. We show that return cross-predictability and
cryptocurrencies' characteristics, including hashing algorithms and proof
types, jointly determine the cryptocurrencies market segmentation. Portfolio
analysis reveals that more centred cryptocurrencies in the network earn higher
risk premiums.
| stat |
Sparse Methods for Automatic Relevance Determination | This work considers methods for imposing sparsity in Bayesian regression with
applications in nonlinear system identification. We first review automatic
relevance determination (ARD) and analytically demonstrate the need to
additional regularization or thresholding to achieve sparse models. We then
discuss two classes of methods, regularization based and thresholding based,
which build on ARD to learn parsimonious solutions to linear problems. In the
case of orthogonal covariates, we analytically demonstrate favorable
performance with regards to learning a small set of active terms in a linear
system with a sparse solution. Several example problems are presented to
compare the set of proposed methods in terms of advantages and limitations to
ARD in bases with hundreds of elements. The aim of this paper is to analyze and
understand the assumptions that lead to several algorithms and to provide
theoretical and empirical results so that the reader may gain insight and make
more informed choices regarding sparse Bayesian regression.
| stat |
Latent Causal Socioeconomic Health Index | This research develops a model-based LAtent Causal Socioeconomic Health
(LACSH) index at the national level. We build upon the latent health factor
index (LHFI) approach that has been used to assess the unobservable
ecological/ecosystem health. This framework integratively models the
relationship between metrics, the latent health, and the covariates that drive
the notion of health. In this paper, the LHFI structure is integrated with
spatial modeling and statistical causal modeling, so as to evaluate the impact
of a continuous policy variable (mandatory maternity leave days and
government's expenditure on healthcare, respectively) on a nation's
socioeconomic health, while formally accounting for spatial dependency among
the nations. A novel visualization technique for evaluating covariate balance
is also introduced for the case of a continuous policy (treatment) variable. We
apply our LACSH model to countries around the world using data on various
metrics and potential covariates pertaining to different aspects of societal
health. The approach is structured in a Bayesian hierarchical framework and
results are obtained by Markov chain Monte Carlo techniques.
| stat |
Technical Issues in the Interpretation of S-values and Their Relation to
Other Information Measures | An extended technical discussion of $S$-values and unconditional information
can be found in Greenland, 2019. Here we briefly cover several technical topics
mentioned in our main paper, Rafi & Greenland, 2020: Different units for
(scaling of) the $S$-value besides base-2 logs (bits); the importance of
uniformity (validity) of the $P$-value for interpretation of the $S$-value; and
the relation of the $S$-value to other measures of statistical information
about a test hypothesis or model.
| stat |
Hamiltonian Monte Carlo in Inverse Problems; Ill-Conditioning and
Multi-Modality | The Hamiltonian Monte Carlo (HMC) method allows sampling from continuous
densities. Favorable scaling with dimension has led to wide adoption of HMC by
the statistics community. Modern auto-differentiating software should allow
more widespread usage in Bayesian inverse problems. This paper analyzes the two
major difficulties encountered using HMC for inverse problems: poor
conditioning and multi-modality. Novel results on preconditioning and replica
exchange Monte Carlo parameter selection are presented in the context of
spectroscopy. Recommendations are analyzed rigorously in the Gaussian case, and
shown to generalize in a fusion plasma reconstruction.
| stat |
The Nu Class of Low-Degree-Truncated, Rational, Generalized Functions.
Ic. IMSPE-optimal designs with circular-disk prediction domains | This paper is an extension of Part I of a series about Nu-class
multifunctions. A method is presented for computing the integrated mean-squared
prediction error (IMSPE) in the design of computer experiments, when the
prediction domain is a circular disk. The method is extensible to more than two
factors, to prediction domains other than squares or disks, and to a variety of
assumed covariance functions. Three example optimal designs, under Gaussian
covariance with known hyperparameters, are found using the method: an n=1
design centered on the disk; an n=2 continuously rotatable design with assumed
inversion symmetry about the center of the disk; and an n=4 twin-point design
similar to the n=4 twin-point design observed previously for a square
prediction domain [1]. The four-point design on the disk demonstrates that
non-round boundaries are not a prerequisite for the occurrence of twin-point
optimal designs.
| stat |
Handling the Positive-Definite Constraint in the Bayesian Learning Rule | The Bayesian learning rule is a natural-gradient variational inference
method, which not only contains many existing learning algorithms as special
cases but also enables the design of new algorithms. Unfortunately, when
variational parameters lie in an open constraint set, the rule may not satisfy
the constraint and requires line-searches which could slow down the algorithm.
In this work, we address this issue for positive-definite constraints by
proposing an improved rule that naturally handles the constraints. Our
modification is obtained by using Riemannian gradient methods, and is valid
when the approximation attains a \emph{block-coordinate natural
parameterization} (e.g., Gaussian distributions and their mixtures). We propose
a principled way to derive Riemannian gradients and retractions from scratch.
Our method outperforms existing methods without any significant increase in
computation. Our work makes it easier to apply the rule in the presence of
positive-definite constraints in parameter spaces.
| stat |
VIB is Half Bayes | In discriminative settings such as regression and classification there are
two random variables at play, the inputs X and the targets Y. Here, we
demonstrate that the Variational Information Bottleneck can be viewed as a
compromise between fully empirical and fully Bayesian objectives, attempting to
minimize the risks due to finite sampling of Y only. We argue that this
approach provides some of the benefits of Bayes while requiring only some of
the work.
| stat |
Consistent estimation of high-dimensional factor models when the factor
number is over-estimated | A high-dimensional $r$-factor model for an $n$-dimensional vector time series
is characterised by the presence of a large eigengap (increasing with $n$)
between the $r$-th and the $(r+1)$-th largest eigenvalues of the covariance
matrix. Consequently, Principal Component (PC) analysis is the most popular
estimation method for factor models and its consistency, when $r$ is correctly
estimated, is well-established in the literature. However, popular factor
number estimators often suffer from the lack of an obvious eigengap in
empirical eigenvalues and tend to over-estimate $r$ due, for example, to the
existence of non-pervasive factors affecting only a subset of the series. We
show that the errors in the PC estimators resulting from the over-estimation of
$r$ are non-negligible, which in turn lead to the violation of the conditions
required for factor-based large covariance estimation. To remedy this, we
propose new estimators of the factor model based on scaling the entries of the
sample eigenvectors. We show both theoretically and numerically that the
proposed estimators successfully control for the over-estimation error, and
investigate their performance when applied to risk minimisation of a portfolio
of financial time series.
| stat |
High-Dimensional Differentially-Private EM Algorithm: Methods and
Near-Optimal Statistical Guarantees | In this paper, we develop a general framework to design differentially
private expectation-maximization (EM) algorithms in high-dimensional latent
variable models, based on the noisy iterative hard-thresholding. We derive the
statistical guarantees of the proposed framework and apply it to three specific
models: Gaussian mixture, mixture of regression, and regression with missing
covariates. In each model, we establish the near-optimal rate of convergence
with differential privacy constraints, and show the proposed algorithm is
minimax rate optimal up to logarithm factors. The technical tools developed for
the high-dimensional setting are then extended to the classic low-dimensional
latent variable models, and we propose a near rate-optimal EM algorithm with
differential privacy guarantees in this setting. Simulation studies and real
data analysis are conducted to support our results.
| stat |
Quantifying Gait Changes Using Microsoft Kinect and Sample Entropy | This study describes a method to quantify potential gait changes in human
subjects. Microsoft Kinect devices were used to provide and track coordinates
of fifteen different joints of a subject over time. Three male subjects walk a
10-foot path multiple times with and without motion-restricting devices. Their
walking patterns were recorded via two Kinect devices through frontal and
sagittal planes. A modified sample entropy (SE) value was computed to quantify
the variability of the time series for each joint. The SE values with and
without motion-restricting devices were used to compare the changes in each
joint. The preliminary results of the experiments show that the proposed
quantification method can detect differences in walking patterns with and
without motion-restricting devices. The proposed method has the potential to be
applied to track personal progress in physical therapy sessions.
| stat |
Efficient parameter sampling for Markov jump processes | Markov jump processes (MJPs) are continuous-time stochastic processes widely
used in a variety of applied disciplines. Inference for MJPs typically proceeds
via Markov chain Monte Carlo, the state-of-the-art being a uniformization-based
auxiliary variable Gibbs sampler. This was designed for situations where the
MJP parameters are known, and Bayesian inference over unknown parameters is
typically carried out by incorporating it into a larger Gibbs sampler. This
strategy of sampling parameters given path, and path given parameters can
result in poor Markov chain mixing. In this work, we propose a simple and
elegant algorithm to address this problem. Our scheme brings
Metropolis-Hastings approaches for discrete-time hidden Markov models to the
continuous-time setting, resulting in a complete and clean recipe for parameter
and path inference in MJPs. In our experiments, we demonstrate superior
performance over Gibbs sampling, as well as another popular approach, particle
MCMC. We also show our sampler inherits geometric mixing from an `ideal'
sampler that operates without computational constraints.
| stat |
End-to-End Learning from Complex Multigraphs with Latent-Graph
Convolutional Networks | We study the problem of end-to-end learning from complex multigraphs with
potentially very large numbers of edges between two vertices, each edge labeled
with rich information. Examples range from communication networks to flights
between airports or financial transaction graphs. We propose Latent-Graph
Convolutional Networks (L-GCNs), which propagate information from these complex
edges to a latent adjacency tensor, after which further downstream tasks can be
performed, such as node classification. We evaluate the performance of several
variations of the model on two synthetic datasets simulating fraud in financial
transaction networks, ensuring the model must make use of edge labels in order
to achieve good classification performance. We find that allowing for nonlinear
interactions on a per-neighbor basis boosts performance significantly, while
showing promising results in an inductive setting. Finally, we demonstrate the
use of L-GCNs on real-world data in the form of an urban transportation
network.
| stat |
Person as Population: A Longitudinal View of Single-Subject Causal
Inference for Analyzing Self-Tracked Health Data | Single-subject health data are becoming increasingly available thanks to
advances in self-tracking technology (e.g., wearable devices, mobile apps,
sensors, implants). Many users and health caregivers seek to use such
observational time series data to recommend changing health practices in order
to achieve desired health outcomes. However, there are few available causal
inference approaches that are flexible enough to analyze such idiographic data.
We develop a recently introduced causal-analysis framework based on n-of-1
randomized trials, and implement a flexible random-forests g-formula approach
to estimating a recurring individualized effect called the "average period
treatment effect". In the process, we argue that our approach essentially
resembles that of a longitudinal study by partitioning a single time series
into periods taking on binary treatment levels. We analyze six years of the
author's own self-tracked physical activity and weight data to demonstrate our
approach, and compare the results of our analysis to one that does not properly
account for confounding.
| stat |
Numerical issues in maximum likelihood parameter estimation for Gaussian
process regression | This article focuses on numerical issues in maximum likelihood parameter
estimation for Gaussian process regression (GPR). This article investigates the
origin of the numerical issues and provides simple but effective improvement
strategies. This work targets a basic problem but a host of studies,
particularly in the literature of Bayesian optimization, rely on off-the-shelf
GPR implementations. For the conclusions of these studies to be reliable and
reproducible, robust GPR implementations are critical.
| stat |
Consistency of random-walk based network embedding algorithms | Random-walk based network embedding algorithms like node2vec and DeepWalk are
widely used to obtain Euclidean representation of the nodes in a network prior
to performing down-stream network inference tasks. Nevertheless, despite their
impressive empirical performance, there is a lack of theoretical results
explaining their behavior. In this paper we studied the node2vec and DeepWalk
algorithms through the perspective of matrix factorization. We analyze these
algorithms in the setting of community detection for stochastic blockmodel
graphs; in particular we established large-sample error bounds and prove
consistent community recovery of node2vec/DeepWalk embedding followed by
k-means clustering. Our theoretical results indicate a subtle interplay between
the sparsity of the observed networks, the window sizes of the random walks,
and the convergence rates of the node2vec/DeepWalk embedding toward the
embedding of the true but unknown edge probabilities matrix. More specifically,
as the network becomes sparser, our results suggest using larger window sizes,
or equivalently, taking longer random walks, in order to attain better
convergence rate for the resulting embeddings. The paper includes numerical
experiments corroborating these observations.
| stat |
Nonparametric Bayesian multi-armed bandits for single cell experiment
design | The problem of maximizing cell type discovery under budget constraints is a
fundamental challenge for the collection and analysis of single-cell
RNA-sequencing (scRNA-seq) data. In this paper, we introduce a simple,
computationally efficient, and scalable Bayesian nonparametric sequential
approach to optimize the budget allocation when designing a large scale
experiment for the collection of scRNA-seq data for the purpose of, but not
limited to, creating cell atlases. Our approach relies on the following tools:
i) a hierarchical Pitman-Yor prior that recapitulates biological assumptions
regarding cellular differentiation, and ii) a Thompson sampling multi-armed
bandit strategy that balances exploitation and exploration to prioritize
experiments across a sequence of trials. Posterior inference is performed by
using a sequential Monte Carlo approach, which allows us to fully exploit the
sequential nature of our species sampling problem. We empirically show that our
approach outperforms state-of-the-art methods and achieves near-Oracle
performance on simulated and scRNA-seq data alike. HPY-TS code is available at
https://github.com/fedfer/HPYsinglecell.
| stat |
Towards Task and Architecture-Independent Generalization Gap Predictors | Can we use deep learning to predict when deep learning works? Our results
suggest the affirmative. We created a dataset by training 13,500 neural
networks with different architectures, on different variations of spiral
datasets, and using different optimization parameters. We used this dataset to
train task-independent and architecture-independent generalization gap
predictors for those neural networks. We extend Jiang et al. (2018) to also use
DNNs and RNNs and show that they outperform the linear model, obtaining
$R^2=0.965$. We also show results for architecture-independent,
task-independent, and out-of-distribution generalization gap prediction tasks.
Both DNNs and RNNs consistently and significantly outperform linear models,
with RNNs obtaining $R^2=0.584$.
| stat |
Hurricane-blackout-heatwave Compound Hazard Risk and Resilience in a
Changing Climate | Hurricanes have caused power outages and blackouts, affecting millions of
customers and inducing severe social and economic impacts. The impacts of
hurricane-caused blackouts may worsen due to increased heat extremes and
possibly increased hurricanes under climate change. We apply hurricane and
heatwave projections with power outage and recovery process analysis to
investigate how the emerging hurricane-blackout-heatwave compound hazard may
vary in a changing climate, for Harris County in Texas (including major part of
Houston City) as an example. We find that, under the high-emissions scenario
RCP8.5, the expected percent of customers experiencing at least one
longer-than-5-day hurricane-induced power outage in a 20-year period would
increase significantly from 14% at the end of the 20th century to 44% at the
end of the 21st century in Harris County. The expected percent of customers who
may experience at least one longer-than-5-day heatwave without power (to
provide air conditioning) would increase alarmingly, from 0.8% to 15.5%. These
increases of risk may be largely avoided if the climate is well controlled
under the stringent mitigation scenario RCP2.6. We also reveal that a moderate
enhancement of critical sectors of the distribution network can significantly
improve the resilience of the entire power grid and mitigate the risk of the
future compound hazard. Together these findings suggest that, in addition to
climate mitigation, climate adaptation actions are urgently needed to improve
the resilience of coastal power systems.
| stat |
Adaptive Design of Experiments for Conservative Estimation of Excursion
Sets | We consider the problem of estimating the set of all inputs that leads a
system to some particular behavior. The system is modeled by an
expensive-to-evaluate function, such as a computer experiment, and we are
interested in its excursion set, i.e. the set of points where the function
takes values above or below some prescribed threshold. The objective function
is emulated with a Gaussian Process (GP) model based on an initial design of
experiments enriched with evaluation results at (batch-)sequentially determined
input points. The GP model provides conservative estimates for the excursion
set, which control false positives while minimizing false negatives. We
introduce adaptive strategies that sequentially select new evaluations of the
function by reducing the uncertainty on conservative estimates. Following the
Stepwise Uncertainty Reduction approach we obtain new evaluations by minimizing
adapted criteria. Tractable formulae for the conservative criteria are derived,
which allow more convenient optimization. The method is benchmarked on random
functions generated under the model assumptions in different scenarios of noise
and batch size. We then apply it to a reliability engineering test case.
Overall, the proposed strategy of minimizing false negatives in conservative
estimation achieves competitive performance both in terms of model-based and
model-free indicators.
| stat |
A Composite Quantile Fourier Neural Network for Multi-Step Probabilistic
Forecasting of Nonstationary Univariate Time Series | Point forecasting of univariate time series is a challenging problem with
extensive work having been conducted. However, nonparametric probabilistic
forecasting of time series, such as in the form of quantiles or prediction
intervals is an even more challenging problem. In an effort to expand the
possible forecasting paradigms we devise and explore an extrapolation-based
approach that has not been applied before for probabilistic forecasting. We
present a novel quantile Fourier neural network is for nonparametric
probabilistic forecasting of univariate time series. Multi-step predictions are
provided in the form of composite quantiles using time as the only input to the
model. This effectively is a form of extrapolation based nonlinear quantile
regression applied for forecasting. Experiments are conducted on eight real
world datasets that demonstrate a variety of periodic and aperiodic patterns.
Nine naive and advanced methods are used as benchmarks including quantile
regression neural network, support vector quantile regression, SARIMA, and
exponential smoothing. The obtained empirical results validate the
effectiveness of the proposed method in providing high quality and accurate
probabilistic predictions.
| stat |
Issues arising from benchmarking single-cell RNA sequencing imputation
methods | On June 25th, 2018, Huang et al. published a computational method SAVER on
Nature Methods for imputing dropout gene expression levels in single cell RNA
sequencing (scRNA-seq) data. Huang et al. performed a set of comprehensive
benchmarking analyses, including comparison with the data from RNA fluorescence
in situ hybridization, to demonstrate that SAVER outperformed two existing
scRNA-seq imputation methods, scImpute and MAGIC. However, their computational
analyses were based on semi-synthetic data that the authors had generated
following the Poisson-Gamma model used in the SAVER method. We have therefore
re-examined Huang et al.'s study. We find that the semi-synthetic data have
very different properties from those of real scRNA-seq data and that the cell
clusters used for benchmarking are inconsistent with the cell types labeled by
biologists. We show that a reanalysis based on real scRNA-seq data and grounded
on biological knowledge of cell types leads to different results and
conclusions from those of Huang et al.
| stat |
Stabilizing a Queue Subject to Action-Dependent Server Performance | We consider a discrete-time system comprising a first-come-first-served
queue, a non-preemptive server, and a scheduler that governs the assignment of
tasks in the queue to the server. The server has an availability state that
indicates, at each instant, whether the server is busy working on a task or is
available. In the latter case, if the queue is nonempty, then a task-assignment
control policy implemented by the scheduler either assigns a new task to the
server or allows it to rest. The server also has an integer-valued activity
state that is non-increasing during rest periods, and is non-decreasing
otherwise. An instantaneous service rate function ascribes to each possible
value of the activity state a probability that the server can complete a task
in one time step. For a typical instantaneous service rate function, the
completion probability decreases (server performance worsens) as the activity
state increases. The scheduler policy has access to the queue size and the
entire state of the server.
In this article, we study the problem of designing scheduler policies that
stabilize the queue. We show that stability, whenever viable, can be achieved
by a simple policy that bases its decisions on the availability state, a
threshold applied to the activity state, and a flag that indicates when the
queue is empty. The supremum of the service rates achievable by stabilizing
policies can be determined by a finite search. Our results remain valid even
when the instantaneous service rate function is not monotonic.
| stat |
Budgeted Multi-Objective Optimization with a Focus on the Central Part
of the Pareto Front -- Extended Version | Optimizing nonlinear systems involving expensive computer experiments with
regard to conflicting objectives is a common challenge. When the number of
experiments is severely restricted and/or when the number of objectives
increases, uncovering the whole set of Pareto optimal solutions is out of
reach, even for surrogate-based approaches: the proposed solutions are
sub-optimal or do not cover the front well. As non-compromising optimal
solutions have usually little point in applications, this work restricts the
search to solutions that are close to the Pareto front center. The article
starts by characterizing this center, which is defined for any type of front.
Next, a Bayesian multi-objective optimization method for directing the search
towards it is proposed. Targeting a subset of the Pareto front allows an
improved optimality of the solutions and a better coverage of this zone, which
is our main concern. A criterion for detecting convergence to the center is
described. If the criterion is triggered, a widened central part of the Pareto
front is targeted such that sufficiently accurate convergence to it is
forecasted within the remaining budget. Numerical experiments show how the
resulting algorithm, C-EHI, better locates the central part of the Pareto front
when compared to state-of-the-art Bayesian algorithms.
| stat |
Marginal Singularity, and the Benefits of Labels in Covariate-Shift | We present new minimax results that concisely capture the relative benefits
of source and target labeled data, under covariate-shift. Namely, we show that
the benefits of target labels are controlled by a transfer-exponent $\gamma$
that encodes how singular Q is locally w.r.t. P, and interestingly allows
situations where transfer did not seem possible under previous insights. In
fact, our new minimax analysis - in terms of $\gamma$ - reveals a continuum of
regimes ranging from situations where target labels have little benefit, to
regimes where target labels dramatically improve classification. We then show
that a recently proposed semi-supervised procedure can be extended to adapt to
unknown $\gamma$, and therefore requests labels only when beneficial, while
achieving minimax transfer rates.
| stat |
A Bulirsch-Stoer algorithm using Gaussian processes | In this paper, we treat the problem of evaluating the asymptotic error in a
numerical integration scheme as one with inherent uncertainty. Adding to the
growing field of probabilistic numerics, we show that Gaussian process
regression (GPR) can be embedded into a numerical integration scheme to allow
for (i) robust selection of the adaptive step-size parameter and; (ii)
uncertainty quantification in predictions of putatively converged numerical
solutions. We present two examples of our approach using Richardson's
extrapolation technique and the Bulirsch-Stoer algorithm. In scenarios where
the error-surface is smooth and bounded, our proposed approach can match the
results of the traditional polynomial (parametric) extrapolation methods. In
scenarios where the error surface is not well approximated by a finite-order
polynomial, e.g. in the vicinity of a pole or in the assessment of a chaotic
system, traditional methods can fail, however, the non-parametric GPR approach
demonstrates the potential to continue to furnish reasonable solutions in these
situations.
| stat |
Partial identification and dependence-robust confidence intervals for
capture-recapture surveys | Capture-recapture (CRC) surveys are widely used to estimate the size of a
population whose members cannot be enumerated directly. When $k$ capture
samples are obtained, counts of unit captures in subsets of samples are
represented naturally by a $2^k$ contingency table in which one element -- the
number of individuals appearing in none of the samples -- remains unobserved.
In the absence of additional assumptions, the population size is not
point-identified. Assumptions about independence between samples are often used
to achieve point-identification. However, real-world CRC surveys often use
convenience samples in which independence cannot be guaranteed, and population
size estimates under independence assumptions may lack empirical credibility.
In this work, we apply the theory of partial identification to show that weak
assumptions or qualitative knowledge about the nature of dependence between
samples can be used to characterize a non-trivial set in which the true
population size lies with high probability. We construct confidence sets for
the population size under bounds on pairwise capture probabilities, and bounds
on the highest order interaction term in a log-linear model using two methods:
test inversion bootstrap confidence intervals, and profile likelihood
confidence intervals. We apply these methods to recent survey data to estimate
the number of people who inject drugs in Brussels, Belgium.
| stat |
Reducing Variance with Sample Allocation Based on Expected Response
Rates | Several techniques exist to assess and reduce nonresponse bias, including
propensity models, calibration methods, or post-stratification. These
approaches can only be applied after the data collection, and assume reliable
information regarding unit nonresponse patterns for the entire population. In
this paper, we demonstrate that sample allocation taking into account the
expected response rates (ERR) have advantages in this context. The performance
of ERR allocation is assessed by comparing the variances of estimates obtained
those arising from a classical allocation proportional to size (PS) and then
applying post-stratification. The main theoretical tool is asymptotic
calculations using the delta-method, and these are complemented with extensive
simulations. The main finding is that the ERR allocation leads to lower
variances than the PS allocation, when the response rates are correctly
specified, and also under a wide range of conditions when the response rates
can not be correctly specified in advance.
| stat |
Selecting and ranking individualized treatment rules with unmeasured
confounding | It is common to compare individualized treatment rules based on the value
function, which is the expected potential outcome under the treatment rule.
Although the value function is not point-identified when there is unmeasured
confounding, it still defines a partial order among the treatment rules under
Rosenbaum's sensitivity analysis model. We first consider how to compare two
treatment rules with unmeasured confounding in the single-decision setting and
then use this pairwise test to rank multiple treatment rules. We consider how
to, among many treatment rules, select the best rules and select the rules that
are better than a control rule. The proposed methods are illustrated using two
real examples, one about the benefit of malaria prevention programs to
different age groups and another about the effect of late retirement on senior
health in different gender and occupation groups.
| stat |
An Interval Estimation Approach to Sample Selection Bias | A widespread and largely unaddressed challenge in statistics is that
non-random participation in study samples can bias the estimation of parameters
of interest. To address this problem, we propose a computationally efficient
interval estimator for a class of population parameters encompassing population
means, ordinary least squares and instrumental variables estimands which makes
minimal assumptions about the selection mechanism. Using results from
stochastic programming, we derive valid confidence intervals and hypothesis
tests based on this estimator. In addition, we demonstrate how to tighten the
intervals by incorporating additional constraints based on population-level
information commonly available to researchers, such as survey response rates
and covariate means. We conduct a comprehensive simulation study to evaluate
the finite sample performance of our estimator and conclude with a real data
study on the causal effect of education on income in the highly-selected UK
Biobank cohort. We are able to demonstrate that our method can produce
informative bounds under relatively few population-level auxiliary constraints.
| stat |
Improving predictions of Bayesian neural nets via local linearization | The generalized Gauss-Newton (GGN) approximation is often used to make
practical Bayesian deep learning approaches scalable by replacing a second
order derivative with a product of first order derivatives. In this paper we
argue that the GGN approximation should be understood as a local linearization
of the underlying Bayesian neural network (BNN), which turns the BNN into a
generalized linear model (GLM). Because we use this linearized model for
posterior inference, we should also predict using this modified model instead
of the original one. We refer to this modified predictive as "GLM predictive"
and show that it effectively resolves common underfitting problems of the
Laplace approximation. It extends previous results in this vein to general
likelihoods and has an equivalent Gaussian process formulation, which enables
alternative inference schemes for BNNs in function space. We demonstrate the
effectiveness of our approach on several standard classification datasets as
well as on out-of-distribution detection. We provide an implementation at
https://github.com/AlexImmer/BNN-predictions.
| stat |
Semi-Implicit Generative Model | To combine explicit and implicit generative models, we introduce
semi-implicit generator (SIG) as a flexible hierarchical model that can be
trained in the maximum likelihood framework. Both theoretically and
experimentally, we demonstrate that SIG can generate high quality samples
especially when dealing with multi-modality. By introducing SIG as an unbiased
regularizer to the generative adversarial network (GAN), we show the interplay
between maximum likelihood and adversarial learning can stabilize the
adversarial training, resist the notorious mode collapsing problem of GANs, and
improve the diversity of generated random samples.
| stat |
Randomization-based joint central limit theorem and efficient covariate
adjustment in stratified $2^K$ factorial experiments | Stratified factorial experiments are widely used in industrial engineering,
clinical trials, and social science to measure the joint effects of several
factors on an outcome. Researchers often use a linear model and analysis of
covariance to analyze experimental results; however, few studies have addressed
the validity and robustness of the resulting inferences because assumptions for
a linear model might not be justified by randomization. In this paper, we
establish the finite-population joint central limit theorem for usual
(unadjusted) factorial effect estimators in stratified $2^K$ factorial
experiments. Our theorem is obtained under randomization-based inference
framework, making use of an extension of the vector form of the
Wald--Wolfowitz--Hoeffding theorem for a linear rank statistic. It is robust to
model misspecification, arbitrary numbers of strata, stratum sizes, and
propensity scores across strata. To improve the estimation and inference
efficiency, we propose three covariate adjustment methods and show that under
mild conditions, the resulting covariate-adjusted factorial effect estimators
are consistent, jointly asymptotically normal, and generally more efficient
than the unadjusted estimator. In addition, we propose Neyman-type conservative
estimators for the asymptotic variances to facilitate valid inferences.
Simulation studies demonstrate the benefits of covariate adjustment methods.
Finally, we apply the proposed methods to analyze a real dataset from a
clinical trial to evaluate the effects of the addition of bevacizumab and/or
carboplatin on pathologic complete response rates in patients with stage II to
III triple-negative breast cancer.
| stat |
DIVA: Domain Invariant Variational Autoencoders | We consider the problem of domain generalization, namely, how to learn
representations given data from a set of domains that generalize to data from a
previously unseen domain. We propose the Domain Invariant Variational
Autoencoder (DIVA), a generative model that tackles this problem by learning
three independent latent subspaces, one for the domain, one for the class, and
one for any residual variations. We highlight that due to the generative nature
of our model we can also incorporate unlabeled data from known or previously
unseen domains. To the best of our knowledge this has not been done before in a
domain generalization setting. This property is highly desirable in fields like
medical imaging where labeled data is scarce. We experimentally evaluate our
model on the rotated MNIST benchmark and a malaria cell images dataset where we
show that (i) the learned subspaces are indeed complementary to each other,
(ii) we improve upon recent works on this task and (iii) incorporating
unlabelled data can boost the performance even further.
| stat |
Conformity bias in the cultural transmission of music sampling
traditions | One of the fundamental questions of cultural evolutionary research is how
individual-level processes scale up to generate population-level patterns.
Previous studies in music have revealed that frequency-based bias (e.g.
conformity and novelty) drives large-scale cultural diversity in different ways
across domains and levels of analysis. Music sampling is an ideal research
model for this process because samples are known to be culturally transmitted
between collaborating artists, and sampling events are reliably documented in
online databases. The aim of the current study was to determine whether
frequency-based bias has played a role in the cultural transmission of music
sampling traditions, using a longitudinal dataset of sampling events across
three decades. Firstly, we assessed whether turn-over rates of popular samples
differ from those expected under neutral evolution. Next, we used agent-based
simulations in an approximate Bayesian computation framework to infer what
level of frequency-based bias likely generated the observed data. Despite
anecdotal evidence of novelty bias, we found that sampling patterns at the
population-level are most consistent with conformity bias.
| stat |
Confidence intervals of prediction accuracy measures for multivariable
prediction models based on the bootstrap-based optimism correction methods | In assessing prediction accuracy of multivariable prediction models, optimism
corrections are essential for preventing biased results. However, in most
published papers of clinical prediction models, the point estimates of the
prediction accuracy measures are corrected by adequate bootstrap-based
correction methods, but their confidence intervals are not corrected, e.g., the
DeLong's confidence interval is usually used for assessing the C-statistic.
These naive methods do not adjust for the optimism bias and do not account for
statistical variability in the estimation of parameters in the prediction
models. Therefore, their coverage probabilities of the true value of the
prediction accuracy measure can be seriously below the nominal level (e.g.,
95%). In this article, we provide two generic bootstrap methods, namely (1)
location-shifted bootstrap confidence intervals and (2) two-stage bootstrap
confidence intervals, that can be generally applied to the bootstrap-based
optimism correction methods, i.e., the Harrell's bias correction, 0.632, and
0.632+ methods. In addition, they can be widely applied to various methods for
prediction model development involving modern shrinkage methods such as the
ridge and lasso regressions. Through numerical evaluations by simulations, the
proposed confidence intervals showed favourable coverage performances. Besides,
the current standard practices based on the optimism-uncorrected methods showed
serious undercoverage properties. To avoid erroneous results, the
optimism-uncorrected confidence intervals should not be used in practice, and
the adjusted methods are recommended instead. We also developed the R package
predboot for implementing these methods (https://github.com/nomahi/predboot).
The effectiveness of the proposed methods are illustrated via applications to
the GUSTO-I clinical trial.
| stat |
Bayesian model for early dose-finding in phase I trials with multiple
treatment courses | Dose-finding clinical trials in oncology aim to determine the maximum
tolerated dose (MTD) of a new drug, generally defined by the proportion of
patients with short-term dose-limiting toxicities (DLTs). Model-based
approaches for such phase I oncology trials have been widely designed and are
mostly restricted to the DLTs occurring during the first cycle of treatment,
although patients continue to receive treatment for multiple cycles. We aim to
estimate the probability of DLTs over sequences of treatment cycles via a
Bayesian cumulative modeling approach, where the probability of DLT is modeled
taking into account the cumulative effect of the administered drug and the DLT
cycle of occurrence. We propose a design, called DICE (Dose-fInding
CumulativE), for dose escalation and de-escalation according to previously
observed toxicities, which aims at finding the MTD sequence (MTS). We performed
an extensive simulation study comparing this approach to the time-to-event
continual reassessment method (TITE-CRM) and to a benchmark. In general, our
approach achieved a better or comparable percentage of correct MTS selection.
Moreover, we investigated the DICE prediction ability.
| stat |
Greater Than the Sum of its Parts: Computationally Flexible Bayesian
Hierarchical Modeling | We propose a multistage method for making inference at all levels of a
Bayesian hierarchical model (BHM) using natural data partitions to increase
efficiency by allowing computations to take place in parallel form using
software that is most appropriate for each data partition. The full
hierarchical model is then approximated by the product of independent normal
distributions for the data component of the model. In the second stage, the
Bayesian maximum {\it a posteriori} (MAP) estimator is found by maximizing the
approximated posterior density with respect to the parameters. If the
parameters of the model can be represented as normally distributed random
effects then the second stage optimization is equivalent to fitting a
multivariate normal linear mixed model. This method can be extended to account
for common fixed parameters shared between data partitions, as well as
parameters that are distinct between partitions. In the case of distinct
parameter estimation, we consider a third stage that re-estimates the distinct
parameters for each data partition based on the results of the second stage.
This allows more information from the entire data set to properly inform the
posterior distributions of the distinct parameters. The method is demonstrated
with two ecological data sets and models, a random effects GLM and an
Integrated Population Model (IPM). The multistage results were compared to
estimates from models fit in single stages to the entire data set. Both
examples demonstrate that multistage point and posterior standard deviation
estimates closely approximate those obtained from fitting the models with all
data simultaneously and can therefore be considered for fitting hierarchical
Bayesian models when it is computationally prohibitive to do so in one step.
| stat |
The continuous Bernoulli: fixing a pervasive error in variational
autoencoders | Variational autoencoders (VAE) have quickly become a central tool in machine
learning, applicable to a broad range of data types and latent variable models.
By far the most common first step, taken by seminal papers and by core software
libraries alike, is to model MNIST data using a deep network parameterizing a
Bernoulli likelihood. This practice contains what appears to be and what is
often set aside as a minor inconvenience: the pixel data is [0,1] valued, not
{0,1} as supported by the Bernoulli likelihood. Here we show that, far from
being a triviality or nuisance that is convenient to ignore, this error has
profound importance to VAE, both qualitative and quantitative. We introduce and
fully characterize a new [0,1]-supported, single parameter distribution: the
continuous Bernoulli, which patches this pervasive bug in VAE. This
distribution is not nitpicking; it produces meaningful performance improvements
across a range of metrics and datasets, including sharper image samples, and
suggests a broader class of performant VAE.
| stat |
The Bayesian Spatial Bradley--Terry Model: Urban Deprivation Modeling in
Tanzania | Identifying the most deprived regions of any country or city is key if policy
makers are to design successful interventions. However, locating areas with the
greatest need is often surprisingly challenging in developing countries. Due to
the logistical challenges of traditional household surveying, official
statistics can be slow to be updated; estimates that exist can be coarse, a
consequence of prohibitive costs and poor infrastructures; and mass
urbanisation can render manually surveyed figures rapidly out-of-date.
Comparative judgement models, such as the Bradley--Terry model, offer a
promising solution. Leveraging local knowledge, elicited via comparisons of
different areas' affluence, such models can both simplify logistics and
circumvent biases inherent to house-hold surveys. Yet widespread adoption
remains limited, due to the large amount of data existing approaches still
require. We address this via development of a novel Bayesian Spatial
Bradley--Terry model, which substantially decreases the amount of data
comparisons required for effective inference. This model integrates a network
representation of the city or country, along with assumptions of spatial
smoothness that allow deprivation in one area to be informed by neighbouring
areas. We demonstrate the practical effectiveness of this method, through a
novel comparative judgement data set collected in Dar es Salaam, Tanzania.
| stat |
Structure Learning of Gaussian Markov Random Fields with False Discovery
Rate Control | In this paper, we propose a new estimation procedure for discovering the
structure of Gaussian Markov random fields (MRFs) with false discovery rate
(FDR) control, making use of the sorted l1-norm (SL1) regularization. A
Gaussian MRF is an acyclic graph representing a multivariate Gaussian
distribution, where nodes are random variables and edges represent the
conditional dependence between the connected nodes. Since it is possible to
learn the edge structure of Gaussian MRFs directly from data, Gaussian MRFs
provide an excellent way to understand complex data by revealing the dependence
structure among many inputs features, such as genes, sensors, users, documents,
etc. In learning the graphical structure of Gaussian MRFs, it is desired to
discover the actual edges of the underlying but unknown probabilistic graphical
model-it becomes more complicated when the number of random variables
(features) p increases, compared to the number of data points n. In particular,
when p >> n, it is statistically unavoidable for any estimation procedure to
include false edges. Therefore, there have been many trials to reduce the false
detection of edges, in particular, using different types of regularization on
the learning parameters. Our method makes use of the SL1 regularization,
introduced recently for model selection in linear regression. We focus on the
benefit of SL1 regularization that it can be used to control the FDR of
detecting important random variables. Adapting SL1 for probabilistic graphical
models, we show that SL1 can be used for the structure learning of Gaussian
MRFs using our suggested procedure nsSLOPE (neighborhood selection Sorted L-One
Penalized Estimation), controlling the FDR of detecting edges.
| stat |
Parametric generation of conditional geological realizations using
generative neural networks | Deep learning techniques are increasingly being considered for geological
applications where -- much like in computer vision -- the challenges are
characterized by high-dimensional spatial data dominated by multipoint
statistics. In particular, a novel technique called generative adversarial
networks has been recently studied for geological parametrization and
synthesis, obtaining very impressive results that are at least qualitatively
competitive with previous methods. The method obtains a neural network
parametrization of the geology -- so-called a generator -- that is capable of
reproducing very complex geological patterns with dimensionality reduction of
several orders of magnitude. Subsequent works have addressed the conditioning
task, i.e. using the generator to generate realizations honoring spatial
observations (hard data). The current approaches, however, do not provide a
parametrization of the conditional generation process. In this work, we propose
a method to obtain a parametrization for direct generation of conditional
realizations. The main idea is to simply extend the existing generator network
by stacking a second inference network that learns to perform the conditioning.
This inference network is a neural network trained to sample a posterior
distribution derived using a Bayesian formulation of the conditioning task. The
resulting extended neural network thus provides the conditional
parametrization. Our method is assessed on a benchmark image of binary
channelized subsurface, obtaining very promising results for a wide variety of
conditioning configurations.
| stat |
Controlled Sequential Monte Carlo | Sequential Monte Carlo methods, also known as particle methods, are a popular
set of techniques for approximating high-dimensional probability distributions
and their normalizing constants. These methods have found numerous applications
in statistics and related fields; e.g. for inference in non-linear non-Gaussian
state space models, and in complex static models. Like many Monte Carlo
sampling schemes, they rely on proposal distributions which crucially impact
their performance. We introduce here a class of controlled sequential Monte
Carlo algorithms, where the proposal distributions are determined by
approximating the solution to an associated optimal control problem using an
iterative scheme. This method builds upon a number of existing algorithms in
econometrics, physics, and statistics for inference in state space models, and
generalizes these methods so as to accommodate complex static models. We
provide a theoretical analysis concerning the fluctuation and stability of this
methodology that also provides insight into the properties of related
algorithms. We demonstrate significant gains over state-of-the-art methods at a
fixed computational complexity on a variety of applications.
| stat |
The leave-one-covariate-out conditional randomization test | Conditional independence testing is an important problem, yet provably hard
without assumptions. One of the assumptions that has become popular of late is
called "model-X", where we assume we know the joint distribution of the
covariates, but assume nothing about the conditional distribution of the
outcome given the covariates. Knockoffs is a popular methodology associated
with this framework, but it suffers from two main drawbacks: only one-bit
$p$-values are available for inference on each variable, and the method is
randomized with significant variability across runs in practice. The
conditional randomization test (CRT) is thought to be the "right" solution
under model-X, but usually viewed as computationally inefficient. This paper
proposes a computationally efficient leave-one-covariate-out (LOCO) CRT that
addresses both drawbacks of knockoffs. LOCO CRT produces valid $p$-values that
can be used to control the familywise error rate, and has nearly zero
algorithmic variability. For L1 regularized M-estimators, we develop an even
faster variant called L1ME CRT, which reuses computation by leveraging a novel
observation about the stability of the cross-validated lasso to removing
inactive variables. Last, for multivariate Gaussian covariates, we present a
closed form expression for the LOCO CRT $p$-value, thus completely eliminating
resampling in this important special case.
| stat |
Detecting and classifying moments in basketball matches using sensor
tracked data | Data analytics in sports is crucial to evaluate the performance of single
players and the whole team. The literature proposes a number of tools for both
offence and defence scenarios. Data coming from tracking location of players,
in this respect, may be used to enrich the amount of useful information. In
basketball, however, actions are interleaved with inactive periods. This paper
describes a methodological approach to automatically identify active periods
during a game and to classify them as offensive or defensive. The method is
based on the application of thresholds to players kinematic parameters, whose
values undergo a tuning strategy similar to Receiver Operating Characteristic
curves, using a ground truth extracted from the video of the games.
| stat |
Bayesian Optimization of Composite Functions | We consider optimization of composite objective functions, i.e., of the form
$f(x)=g(h(x))$, where $h$ is a black-box derivative-free expensive-to-evaluate
function with vector-valued outputs, and $g$ is a cheap-to-evaluate real-valued
function. While these problems can be solved with standard Bayesian
optimization, we propose a novel approach that exploits the composite structure
of the objective function to substantially improve sampling efficiency. Our
approach models $h$ using a multi-output Gaussian process and chooses where to
sample using the expected improvement evaluated on the implied non-Gaussian
posterior on $f$, which we call expected improvement for composite functions
(\ei). Although \ei\ cannot be computed in closed form, we provide a novel
stochastic gradient estimator that allows its efficient maximization. We also
show that our approach is asymptotically consistent, i.e., that it recovers a
globally optimal solution as sampling effort grows to infinity, generalizing
previous convergence results for classical expected improvement. Numerical
experiments show that our approach dramatically outperforms standard Bayesian
optimization benchmarks, reducing simple regret by several orders of magnitude.
| stat |
A Unifying Framework for Variance Reduction Algorithms for Finding
Zeroes of Monotone Operators | It is common to encounter large-scale monotone inclusion problems where the
objective has a finite sum structure. We develop a general framework for
variance-reduced forward-backward splitting algorithms for this problem. This
framework includes a number of existing deterministic and variance-reduced
algorithms for function minimization as special cases, and it is also
applicable to more general problems such as saddle-point problems and
variational inequalities. With a carefully constructed Lyapunov function, we
show that the algorithms covered by our framework enjoy a linear convergence
rate in expectation under mild assumptions. We further consider Catalyst
acceleration and asynchronous implementation to reduce the algorithmic
complexity and computation time. We apply our proposed framework to a policy
evaluation problem and a strongly monotone two-player game, both of which fall
outside of function minimization.
| stat |
Distributional Gaussian Process Layers for Outlier Detection in Image
Segmentation | We propose a parameter efficient Bayesian layer for hierarchical
convolutional Gaussian Processes that incorporates Gaussian Processes operating
in Wasserstein-2 space to reliably propagate uncertainty. This directly
replaces convolving Gaussian Processes with a distance-preserving affine
operator on distributions. Our experiments on brain tissue-segmentation show
that the resulting architecture approaches the performance of well-established
deterministic segmentation algorithms (U-Net), which has never been achieved
with previous hierarchical Gaussian Processes. Moreover, by applying the same
segmentation model to out-of-distribution data (i.e., images with pathology
such as brain tumors), we show that our uncertainty estimates result in
out-of-distribution detection that outperforms the capabilities of previous
Bayesian networks and reconstruction-based approaches that learn normative
distributions.
| stat |
Independent Component Analysis for Compositional Data | Compositional data represent a specific family of multivariate data, where
the information of interest is contained in the ratios between parts rather
than in absolute values of single parts. The analysis of such specific data is
challenging as the application of standard multivariate analysis tools on the
raw observations can lead to spurious results. Hence, it is appropriate to
apply certain transformations prior further analysis. One popular multivariate
data analysis tool is independent component analysis. Independent component
analysis aims to find statistically independent components in the data and as
such might be seen as an extension to principal component analysis. In this
paper we examine an approach of how to apply independent component analysis on
compositional data by respecting the nature of the former and demonstrate the
usefulness of this procedure on a metabolomic data set.
| stat |
Information geometry for approximate Bayesian computation | The goal of this paper is to explore the basic Approximate Bayesian
Computation (ABC) algorithm via the lens of information theory. ABC is a widely
used algorithm in cases where the likelihood of the data is hard to work with
or intractable, but one can simulate from it. We use relative entropy ideas to
analyze the behavior of the algorithm as a function of the threshold parameter
and of the size of the data. Relative entropy here is data driven as it depends
on the values of the observed statistics. Relative entropy also allows us to
explore the effect of the distance metric and sets up a mathematical framework
for sensitivity analysis allowing to find important directions which could lead
to lower computational cost of the algorithm for the same level of accuracy. In
addition, we also investigate the bias of the estimators for generic
observables as a function of both the threshold parameters and the size of the
data. Our analysis provides error bounds on performance for positive tolerances
and finite sample sizes. Simulation studies complement and illustrate the
theoretical results.
| stat |
Bayesian Dynamic Modeling and Monitoring of Network Flows | In the context of a motivating study of dynamic network flow data on a
large-scale e-commerce web site, we develop Bayesian models for
on-line/sequential analysis for monitoring and adapting to changes reflected in
node-node traffic. For large-scale networks, we customize core Bayesian time
series analysis methods using dynamic generalized linear models (DGLMs). These
are integrated into the context of multivariate networks using the concept of
decouple/recouple that was recently introduced in multivariate time series.
This method enables flexible dynamic modeling of flows on large-scale networks
and exploitation of partial parallelization of analysis while maintaining
coherence with an over-arching multivariate dynamic flow model. This approach
is anchored in a case-study on internet data, with flows of visitors to a
commercial news web site defining a long time series of node-node counts on
over 56,000 node pairs. Central questions include characterizing inherent
stochasticity in traffic patterns, understanding node-node interactions,
adapting to dynamic changes in flows and allowing for sensitive monitoring to
flag anomalies. The methodology of dynamic network DGLMs applies to many
dynamic network flow studies.
| stat |
A Longitudinal Framework for Predicting Nonresponse in Panel Surveys | Nonresponse in panel studies can lead to a substantial loss in data quality
due to its potential to introduce bias and distort survey estimates. Recent
work investigates the usage of machine learning to predict nonresponse in
advance, such that predicted nonresponse propensities can be used to inform the
data collection process. However, predicting nonresponse in panel studies
requires accounting for the longitudinal data structure in terms of model
building, tuning, and evaluation. This study proposes a longitudinal framework
for predicting nonresponse with machine learning and multiple panel waves and
illustrates its application. With respect to model building, this approach
utilizes information from multiple waves by introducing features that aggregate
previous (non)response patterns. Concerning model tuning and evaluation,
temporal cross-validation is employed by iterating through pairs of panel waves
such that the training and test sets move in time. Implementing this approach
with data from a German probability-based mixed-mode panel shows that
aggregating information over multiple panel waves can be used to build
prediction models with competitive and robust performance over all test waves.
| stat |
Minimum Detectable Effect Size Computations for Cluster-Level Regression
Discontinuity: Quadratic Functional Form and Beyond | This study extends power formulas proposed by Schochet (2008) assuming that
the cluster-level score variable follows quadratic functional form. Results
reveal that we need not be concerned with treatment by linear term interaction,
and polynomial degree up to second order for symmetric truncation intervals. In
comparison, every slight change in the functional form alters sample size
requirements for asymmetric truncation intervals. Finally, an empirical
framework beyond quadratic functional form is provided when the asymptotic
variance of the treatment effect is untraceable. In this case, the CRD design
effect is either computed from moments of the sample or approximate population
moments via simulation. Formulas for quadratic functional form and the extended
empirical framework are implemented in the cosa R package and companion Shiny
web application.
| stat |
Max-and-Smooth: a two-step approach for approximate Bayesian inference
in latent Gaussian models | With modern high-dimensional data, complex statistical models are necessary,
requiring computationally feasible inference schemes. We introduce
Max-and-Smooth, an approximate Bayesian inference scheme for a flexible class
of latent Gaussian models (LGMs) where one or more of the likelihood parameters
are modeled by latent additive Gaussian processes. Max-and-Smooth consists of
two-steps. In the first step (Max), the likelihood function is approximated by
a Gaussian density with mean and covariance equal to either (a) the maximum
likelihood estimate and the inverse observed information, respectively, or (b)
the mean and covariance of the normalized likelihood function. In the second
step (Smooth), the latent parameters and hyperparameters are inferred and
smoothed with the approximated likelihood function. The proposed method ensures
that the uncertainty from the first step is correctly propagated to the second
step. Since the approximated likelihood function is Gaussian, the approximate
posterior density of the latent parameters of the LGM (conditional on the
hyperparameters) is also Gaussian, thus facilitating efficient posterior
inference in high dimensions. Furthermore, the approximate marginal posterior
distribution of the hyperparameters is tractable, and as a result, the
hyperparameters can be sampled independently of the latent parameters. In the
case of a large number of independent data replicates, sparse precision
matrices, and high-dimensional latent vectors, the speedup is substantial in
comparison to an MCMC scheme that infers the posterior density from the exact
likelihood function. The proposed inference scheme is demonstrated on one
spatially referenced real dataset and on simulated data mimicking spatial,
temporal, and spatio-temporal inference problems. Our results show that
Max-and-Smooth is accurate and fast.
| stat |
Latent Cross-population Dynamic Time-series Analysis of High-dimensional
Neural Recordings | An important problem in analysis of neural data is to characterize
interactions across brain regions from high-dimensional multiple-electrode
recordings during a behavioral experiment. Lead-lag effects indicate possible
directional flows of neural information, but they are often transient,
appearing during short intervals of time. Such non-stationary interactions can
be difficult to identify, but they can be found by taking advantage of the
replication structure inherent to many neurophysiological experiments. To
describe non-stationary interactions between replicated pairs of
high-dimensional time series, we developed a method of estimating latent,
non-stationary cross-correlation. Our approach begins with an extension of
probabilistic CCA to the time series setting, which provides a model-based
interpretation of multiset CCA. Because the covariance matrix describing
non-stationary dependence is high-dimensional, we assume sparsity of
cross-correlations within a range of possible interesting lead-lag effects. We
show that the method can perform well in realistic settings and we apply it to
192 simultaneous local field potential (LFP) recordings from prefrontal cortex
(PFC) and visual cortex (area V4) during a visual memory task. We find lead-lag
relationships that are highly plausible, being consistent with related results
in the literature.
| stat |
Torus Probabilistic Principal Component Analysis | One of the most common problems that any technique encounters is the high
dimensionality of the input data. This yields several problems in the
subsequently statistical methods due to so called "curse of dimensionality".
Several dimension reduction methods have been proposed in the literature, until
now, to accomplish the goal of having smaller dimension space. The most popular
among them, is the so called Principal Component Analysis (PCA). One of the
extensions of PCA is Probabilistic PCA (known as PPCA). In PPCA, a
probabilistic model is used to perform the dimension reduction. By convention,
there are cases in applied sciences, e.g. Bioinformatics, Biology and Geology
that the data at hand are in non Euclidean space. Elaborating further, an
important situation is when each variable poses a direction so that a data
point is lying on a torus. According to its importance, the extension of the
PCA to such data is being under attention during last decades. In this paper,
we introduce a Probabilistic PCA for data on torus (TPPCA), assuming a
Multivariate Wrapped Normal distribution as the underline model. Furthermore,
we present certain appropriate algorithms to compute this dimension reduction
technique and then, we compare our proposal with other dimension reduction
techniques using examples based on real datasets and a small Monte Carlo
simulation.
| stat |
Couplings for Multinomial Hamiltonian Monte Carlo | Hamiltonian Monte Carlo (HMC) is a popular sampling method in Bayesian
inference. Recently, Heng & Jacob (2019) studied Metropolis HMC with couplings
for unbiased Monte Carlo estimation, establishing a generic parallelizable
scheme for HMC. However, in practice a different HMC method, multinomial HMC,
is considered as the go-to method, e.g. as part of the no-U-turn sampler. In
multinomial HMC, proposed states are not limited to end-points as in Metropolis
HMC; instead points along the entire trajectory can be proposed. In this paper,
we establish couplings for multinomial HMC, based on optimal transport for
multinomial sampling in its transition. We prove an upper bound for the meeting
time - the time it takes for the coupled chains to meet - based on the notion
of local contractivity. We evaluate our methods using three targets: 1,000
dimensional Gaussians, logistic regression and log-Gaussian Cox point
processes. Compared to Heng & Jacob (2019), coupled multinomial HMC generally
attains a smaller meeting time, and is more robust to choices of step sizes and
trajectory lengths, which allows re-use of existing adaptation methods for HMC.
These improvements together paves the way for a wider and more practical use of
coupled HMC methods.
| stat |
Constructing Stabilized Dynamic Treatment Regimes for Censored Data | Stabilized dynamic treatment regimes are sequential decision rules for
individual patients that not only adaptive throughout the disease progression
but also remain consistent over time in format. The estimation of stabilized
dynamic treatment regimes becomes more complicated when the clinical outcome of
interest is a survival time subject to censoring. To address this challenge, we
propose two novel methods, censored shared-Q-learning and censored
shared-O-learning. Both methods incorporate clinical preferences into a
qualitative rule, where the parameters indexing the decision rules are shared
across different stages and estimated simultaneously. We use extensive
simulation studies to demonstrate the superior performance of the proposed
methods. The methods are further applied to the Framingham Study to derive
treatment rules for cardiovascular disease.
| stat |
Fast Global Convergence for Low-rank Matrix Recovery via Riemannian
Gradient Descent with Random Initialization | In this paper, we propose a new global analysis framework for a class of
low-rank matrix recovery problems on the Riemannian manifold. We analyze the
global behavior for the Riemannian optimization with random initialization. We
use the Riemannian gradient descent algorithm to minimize a least squares loss
function, and study the asymptotic behavior as well as the exact convergence
rate. We reveal a previously unknown geometric property of the low-rank matrix
manifold, which is the existence of spurious critical points for the simple
least squares function on the manifold. We show that under some assumptions,
the Riemannian gradient descent starting from a random initialization with high
probability avoids these spurious critical points and only converges to the
ground truth in nearly linear convergence rate, i.e.
$\mathcal{O}(\text{log}(\frac{1}{\epsilon})+ \text{log}(n))$ iterations to
reach an $\epsilon$-accurate solution. We use two applications as examples for
our global analysis. The first one is a rank-1 matrix recovery problem. The
second one is a generalization of the Gaussian phase retrieval problem. It only
satisfies the weak isometry property, but has behavior similar to that of the
first one except for an extra saddle set. Our convergence guarantee is nearly
optimal and almost dimension-free, which fully explains the numerical
observations. The global analysis can be potentially extended to other data
problems with random measurement structures and empirical least squares loss
functions.
| stat |
Model-assisted estimation in high-dimensional settings for survey data | Model-assisted estimators have attracted a lot of attention in the last three
decades. These estimators attempt to make an efficient use of auxiliary
information available at the estimation stage. A working model linking the
survey variable to the auxiliary variables is specified and fitted on the
sample data to obtain a set of predictions, which are then incorporated in the
estimation procedures. A nice feature of model-assisted procedures is that they
maintain important design properties such as consistency and asymptotic
unbiasedness irrespective of whether or not the working model is correctly
specified. In this article, we examine several model-assisted estimators from a
design-based point of view and in a high-dimensional setting, including
penalized estimators and tree-based estimators. We conduct an extensive
simulation study using data from the Irish Commission for Energy Regulation
Smart Metering Project, in order to assess the performance of several
model-assisted estimators in terms of bias and efficiency in this
high-dimensional data set.
| stat |
Model Averaging and its Use in Economics | The method of model averaging has become an important tool to deal with model
uncertainty, for example in situations where a large amount of different
theories exist, as are common in economics. Model averaging is a natural and
formal response to model uncertainty in a Bayesian framework, and most of the
paper deals with Bayesian model averaging. The important role of the prior
assumptions in these Bayesian procedures is highlighted. In addition,
frequentist model averaging methods are also discussed. Numerical methods to
implement these methods are explained, and I point the reader to some freely
available computational resources. The main focus is on uncertainty regarding
the choice of covariates in normal linear regression models, but the paper also
covers other, more challenging, settings, with particular emphasis on sampling
models commonly used in economics. Applications of model averaging in economics
are reviewed and discussed in a wide range of areas, among which growth
economics, production modelling, finance and forecasting macroeconomic
quantities.
| stat |
Decomposing spectral and phasic differences in non-linear features
between datasets | When employing non-linear methods to characterise complex systems, it is
important to determine to what extent they are capturing genuine non-linear
phenomena that could not be assessed by simpler spectral methods. Specifically,
we are concerned with the problem of quantifying spectral and phasic effects on
an observed difference in a non-linear feature between two systems (or two
states of the same system). Here we derive, from a sequence of null models, a
decomposition of the difference in an observable into spectral, phasic, and
spectrum-phase interaction components. Our approach makes no assumptions about
the structure of the data and adds nuance to a wide range of time series
analyses.
| stat |
Estimating Barycenters of Measures in High Dimensions | Barycentric averaging is a principled way of summarizing populations of
measures. Existing algorithms for estimating barycenters typically parametrize
them as weighted sums of Diracs and optimize their weights and/or locations.
However, these approaches do not scale to high-dimensional settings due to the
curse of dimensionality. In this paper, we propose a scalable and general
algorithm for estimating barycenters of measures in high dimensions. The key
idea is to turn the optimization over measures into an optimization over
generative models, introducing inductive biases that allow the method to scale
while still accurately estimating barycenters. We prove local convergence under
mild assumptions on the discrepancy showing that the approach is well-posed. We
demonstrate that our method is fast, achieves good performance on
low-dimensional problems, and scales to high-dimensional settings. In
particular, our approach is the first to be used to estimate barycenters in
thousands of dimensions.
| stat |
Joint longitudinal models for dealing with missing at random data in
trial-based economic evaluations | Health economic evaluations based on patient-level data collected alongside
clinical trials~(e.g. health related quality of life and resource use measures)
are an important component of the process which informs resource allocation
decisions. Almost inevitably, the analysis is complicated by the fact that some
individuals drop out from the study, which causes their data to be unobserved
at some time point. Current practice performs the evaluation by handling the
missing data at the level of aggregated variables (e.g. QALYs), which are
obtained by combining the economic data over the duration of the study, and are
often conducted under a missing at random (MAR) assumption. However, this
approach may lead to incorrect inferences since it ignores the longitudinal
nature of the data and may end up discarding a considerable amount of
observations from the analysis. We propose the use of joint longitudinal models
to extend standard cost-effectiveness analysis methods by taking into account
the longitudinal structure and incorporate all available data to improve the
estimation of the targeted quantities under MAR. Our approach is compared to
popular missingness approaches in trial-based analyses, motivated by an
exploratory simulation study, and applied to data from two real case studies.
| stat |
Replica-exchange Nos\'e-Hoover dynamics for Bayesian learning on large
datasets | In this paper, we present a new practical method for Bayesian learning that
can rapidly draw representative samples from complex posterior distributions
with multiple isolated modes in the presence of mini-batch noise. This is
achieved by simulating a collection of replicas in parallel with different
temperatures and periodically swapping them. When evolving the replicas'
states, the Nos\'e-Hoover dynamics is applied, which adaptively neutralizes the
mini-batch noise. To perform proper exchanges, a new protocol is developed with
a noise-aware test of acceptance, by which the detailed balance is reserved in
an asymptotic way. While its efficacy on complex multimodal posteriors has been
illustrated by testing over synthetic distributions, experiments with deep
Bayesian neural networks on large-scale datasets have shown its significant
improvements over strong baselines.
| stat |
Mat\'ern Gaussian processes on Riemannian manifolds | Gaussian processes are an effective model class for learning unknown
functions, particularly in settings where accurately representing predictive
uncertainty is of key importance. Motivated by applications in the physical
sciences, the widely-used Mat\'ern class of Gaussian processes has recently
been generalized to model functions whose domains are Riemannian manifolds, by
re-expressing said processes as solutions of stochastic partial differential
equations. In this work, we propose techniques for computing the kernels of
these processes on compact Riemannian manifolds via spectral theory of the
Laplace-Beltrami operator in a fully constructive manner, thereby allowing them
to be trained via standard scalable techniques such as inducing point methods.
We also extend the generalization from the Mat\'ern to the widely-used squared
exponential Gaussian process. By allowing Riemannian Mat\'ern Gaussian
processes to be trained using well-understood techniques, our work enables
their use in mini-batch, online, and non-conjugate settings, and makes them
more accessible to machine learning practitioners.
| stat |
Model-based clustering and classification using mixtures of multivariate
skewed power exponential distributions | Families of mixtures of multivariate power exponential (MPE) distributions
have been previously introduced and shown to be competitive for cluster
analysis in comparison to other elliptical mixtures including mixtures of
Gaussian distributions. Herein, we propose a family of mixtures of multivariate
skewed power exponential distributions to combine the flexibility of the MPE
distribution with the ability to model skewness. These mixtures are more robust
to variations from normality and can account for skewness, varying tail weight,
and peakedness of data. A generalized expectation-maximization approach
combining minorization-maximization and optimization based on accelerated line
search algorithms on the Stiefel manifold is used for parameter estimation.
These mixtures are implemented both in the model-based clustering and
classification frameworks. Both simulated and benchmark data are used for
illustration and comparison to other mixture families.
| stat |
The Binary Space Partitioning-Tree Process | The Mondrian process represents an elegant and powerful approach for space
partition modelling. However, as it restricts the partitions to be
axis-aligned, its modelling flexibility is limited. In this work, we propose a
self-consistent Binary Space Partitioning (BSP)-Tree process to generalize the
Mondrian process. The BSP-Tree process is an almost surely right continuous
Markov jump process that allows uniformly distributed oblique cuts in a
two-dimensional convex polygon. The BSP-Tree process can also be extended using
a non-uniform probability measure to generate direction differentiated cuts.
The process is also self-consistent, maintaining distributional invariance
under a restricted subdomain. We use Conditional-Sequential Monte Carlo for
inference using the tree structure as the high-dimensional variable. The
BSP-Tree process's performance on synthetic data partitioning and relational
modelling demonstrates clear inferential improvements over the standard
Mondrian process and other related methods.
| stat |
Confidence sequences for sampling without replacement | Many practical tasks involve sampling sequentially without replacement (WoR)
from a finite population of size $N$, in an attempt to estimate some parameter
$\theta^\star$. Accurately quantifying uncertainty throughout this process is a
nontrivial task, but is necessary because it often determines when we stop
collecting samples and confidently report a result. We present a suite of tools
for designing confidence sequences (CS) for $\theta^\star$. A CS is a sequence
of confidence sets $(C_n)_{n=1}^N$, that shrink in size, and all contain
$\theta^\star$ simultaneously with high probability. We present a generic
approach to constructing a frequentist CS using Bayesian tools, based on the
fact that the ratio of a prior to the posterior at the ground truth is a
martingale. We then present Hoeffding- and empirical-Bernstein-type
time-uniform CSs and fixed-time confidence intervals for sampling WoR, which
improve on previous bounds in the literature and explicitly quantify the
benefit of WoR sampling.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.