title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
A Mathematical Assessment of the Isolation Tree Method for Outliers
Detection in Big Data | In this paper, the mathematical analysis of the Isolation Random Forest
Method (IRF Method) for anomaly detection is presented. We show that the IRF
space can be endowed with a probability induced by the Isolation Tree algorithm
(iTree). In this setting, the convergence of the IRF method is proved using the
Law of Large Numbers. A couple of counterexamples are presented to show that
the original method is inconclusive and no quality certificate can be given,
when using it as a means to detect anomalies. Hence, an alternative version of
IRF is proposed, whose mathematical foundation, as well as its limitations, are
fully justified. Finally, numerical experiments are presented to compare the
performance of the classic IRF with the proposed one.
| stat |
Bayesian sequential data assimilation for COVID-19 forecasting | We introduce a Bayesian sequential data assimilation method for COVID-19
forecasting. It is assumed that suitable transmission, epidemic and observation
models are available and previously validated and the transmission and epidemic
models are coded into a dynamical system. The observation model depends on the
dynamical system state variables and parameters, and is cast as a likelihood
function. We elicit prior distributions of the effective population size, the
dynamical system initial conditions and infectious contact rate, and use Markov
Chain Monte Carlo sampling to make inference and prediction of quantities of
interest (QoI) at the onset of the epidemic outbreak. The forecast is
sequentially updated over a sliding window of epidemic records as new data
becomes available. Prior distributions for the state variables at the new
forecasting time are assembled using the dynamical system, calibrated for the
previous forecast. Moreover, changes in the contact rate and effective
population size are naturally introduced through auto-regressive models on the
corresponding parameters. We show our forecasting method's performance using a
SEIR type model and COVID-19 data from several Mexican localities.
| stat |
The false positive risk: a proposal concerning what to do about p values | It is widely acknowledged that the biomedical literature suffer from a
surfeit of false positive results. Part of the reason for this is the
persistence of the myth that observation of a p value less than 0.05 is
sufficient justification to claim that you've made a discovery.
It is hopeless to expect users to change their reliance on p values unless
they are offered an alternative way of judging the reliability of their
conclusions. If the alternative method is to have a chance of being adopted
widely, it will have to be easy to understand and to calculate. One such
proposal is based on calculation of false positive risk.
It is suggested that p values and confidence intervals should continue to be
given, but that they should be supplemented by a single additional number that
conveys the strength of the evidence better than the p value. This number could
be the minimum false positive risk (that calculated on the assumption of a
prior probability of 0.5, the largest value that can be assumed in the absence
of hard prior data). Alternatively one could specify the prior probability that
it would be necessary to believe in order to achieve a false positive risk of,
say, 0.05.
| stat |
Degree irregularity and rank probability bias in network meta-analysis | Network meta-analysis (NMA) is a statistical technique for the comparison of
treatment options. The nodes of the network are the competing treatments and
edges represent comparisons of treatments in trials. Outcomes of Bayesian NMA
include estimates of treatment effects, and the probabilities that each
treatment is ranked best, second best and so on. How exactly network geometry
affects the accuracy and precision of these outcomes is not fully understood.
Here we carry out a simulation study and find that disparity in the number of
trials involving different treatments leads to a systematic bias in estimated
rank probabilities. This bias is associated with an increased variation in the
precision of treatment effect estimates. Using ideas from the theory of complex
networks, we define a measure of `degree irregularity' to quantify asymmetry in
the number of studies involving each treatment. Our simulations indicate that
more regular networks have more precise treatment effect estimates and smaller
bias of rank probabilities. We also find that degree regularity is a better
indicator of NMA quality than both the total number of studies in a network and
the disparity in the number of trials per comparison. These results have
implications for planning future trials. We demonstrate that choosing trials
which reduce the network's irregularity can improve the precision and accuracy
of NMA outcomes.
| stat |
Prevalence of international migration: an alternative for small area
estimation | This paper introduces an alternative procedure for estimating the prevalence
of international migration at the municipal level in Colombia. The new
methodology uses the empirical best linear unbiased predictor based on a
Fay-Herriot model with target and auxiliary variables available from census
studies and from the Demographic and Health Survey. The proposed alternative
produces prevalence estimates which are consistent with sample sizes and
demographic dynamics in Colombia. Additionally, the estimated coefficients of
variation are lower than 20% for municipalities and large
demographically-relevant capital cities and therefore estimates may be
considered as reliable.
| stat |
Multi-Regularization Reconstruction of One-Dimensional $T_2$
Distributions in Magnetic Resonance Relaxometry with a Gaussian Basis | We consider the inverse problem of recovering the probability distribution
function of $T_2$ relaxation times from NMR transverse relaxometry experiments.
This problem is a variant of the inverse Laplace transform and hence ill-posed.
We cast this within the framework of a Gaussian mixture model to obtain a
least-square problem with an $L_2$ regularization term. We propose a new method
for incorporating regularization into the solution; rather than seeking to
replace the native problem with a suitable mathematically close, regularized,
version, we instead augment the native formulation with regularization. We term
this new approach 'multi-regularization'; it avoids the treacherous process of
selecting a single best regularization parameter $\lambda$ and instead permits
incorporation of several degrees of regularization into the solution. We
illustrate the method with extensive simulation results as well as application
to real experimental data.
| stat |
Monitoring SEIRD model parameters using MEWMA for the COVID-19 pandemic
with application to the State of Qatar | During the current COVID-19 pandemic, decision makers are tasked with
implementing and evaluating strategies for both treatment and disease
prevention. In order to make effective decisions, they need to simultaneously
monitor various attributes of the pandemic such as transmission rate and
infection rate for disease prevention, recovery rate which indicates treatment
effectiveness as well as the mortality rate and others. This work presents a
technique for monitoring the pandemic by employing an Susceptible, Exposed,
Infected, Recovered Death model regularly estimated by an augmented particle
Markov chain Monte Carlo scheme in which the posterior distribution samples are
monitored via Multivariate Exponentially Weighted Average process monitoring.
This is illustrated on the COVID-19 data for the State of Qatar.
| stat |
Optimal multiple testing and design in clinical trials | A central goal in designing clinical trials is to find the test that
maximizes power (or equivalently minimizes required sample size) for finding a
true research hypothesis subject to the constraint of type I error. When there
is more than one test, such as in clinical trials with multiple endpoints, the
issues of optimal design and optimal policies become more complex. In this
paper we address the question of how such optimal tests should be defined and
how they can be found. We review different notions of power and how they relate
to study goals, and also consider the requirements of type I error control and
the nature of the policies. This leads us to formulate the optimal policy
problem as an explicit optimization problem with objective and constraints
which describe its specific desiderata. We describe a complete solution for
deriving optimal policies for two hypotheses, which have desired monotonicity
properties, and are computationally simple. For some of the optimization
formulations this yields optimal policies that are identical to existing
policies, such as Hommel's procedure or the procedure of Bittman et al. (2009),
while for others it yields completely novel and more powerful policies than
existing ones. We demonstrate the nature of our novel policies and their
improved power extensively in simulation and on the APEX study (Cohen et al.,
2016).
| stat |
Interpretable Almost Matching Exactly for Causal Inference | We aim to create the highest possible quality of treatment-control matches
for categorical data in the potential outcomes framework. Matching methods are
heavily used in the social sciences due to their interpretability, but most
matching methods do not pass basic sanity checks: they fail when irrelevant
variables are introduced, and tend to be either computationally slow or produce
low-quality matches. The method proposed in this work aims to match units on a
weighted Hamming distance, taking into account the relative importance of the
covariates; the algorithm aims to match units on as many relevant variables as
possible. To do this, the algorithm creates a hierarchy of covariate
combinations on which to match (similar to downward closure), in the process
solving an optimization problem for each unit in order to construct the optimal
matches. The algorithm uses a single dynamic program to solve all of the
optimization problems simultaneously. Notable advantages of our method over
existing matching procedures are its high-quality matches, versatility in
handling different data distributions that may have irrelevant variables, and
ability to handle missing data by matching on as many available covariates as
possible.
| stat |
Modeling the Heterogeneity in COVID-19's Reproductive Number and its
Impact on Predictive Scenarios | The correct evaluation of the reproductive number $R$ for COVID-19 -- which
characterizes the average number of secondary cases generated by each typical
primary case -- is central in the quantification of the potential scope of the
pandemic and the selection of an appropriate course of action. In most models,
$R$ is modeled as a universal constant for the virus across outbreak clusters
and individuals -- effectively averaging out the inherent variability of the
transmission process due to varying individual contact rates, population
densities, demographics, or temporal factors amongst many. Yet, due to the
exponential nature of epidemic growth, the error due to this simplification can
be rapidly amplified and lead to inaccurate predictions and/or risk evaluation.
From the statistical modeling perspective, the magnitude of the impact of this
averaging remains an open question: how can this intrinsic variability be
percolated into epidemic models, and how can its impact on uncertainty
quantification and predictive scenarios be better quantified? In this paper, we
propose to study this question through a Bayesian perspective, creating a
bridge between the agent-based and compartmental approaches commonly used in
the literature. After deriving a Bayesian model that captures at scale the
heterogeneity of a population and environmental conditions, we simulate the
spread of the epidemic as well as the impact of different social distancing
strategies, and highlight the strong impact of this added variability on the
reported results. We base our discussion on both synthetic experiments --
thereby quantifying of the reliability and the magnitude of the effects -- and
real COVID-19 data.
| stat |
Distribution-Free Prediction Sets with Random Effects | We consider the problem of constructing distribution-free prediction sets
when there are random effects. For iid data, prediction sets can be constructed
using the method of conformal prediction. The validity of this prediction set
hinges on the assumption that the data are exchangeable, which is not true when
there are random effects. We extend the conformal method so that it is valid
with random effects. We develop a CDF pooling approach, a single subsampling
approach, and a repeated subsampling approach to construct conformal prediction
sets in unsupervised and supervised settings. We compare these approaches in
terms of coverage and average set size. We recommend the repeated subsampling
approach that constructs a conformal set by sampling one observation from each
distribution multiple times. Simulations show that this approach has the best
balance between coverage and average conformal set size.
| stat |
Ensemble Quantile Classifier | Both the median-based classifier and the quantile-based classifier are useful
for discriminating high-dimensional data with heavy-tailed or skewed inputs.
But these methods are restricted as they assign equal weight to each variable
in an unregularized way. The ensemble quantile classifier is a more flexible
regularized classifier that provides better performance with high-dimensional
data, asymmetric data or when there are many irrelevant extraneous inputs. The
improved performance is demonstrated by a simulation study as well as an
application to text categorization. It is proven that the estimated parameters
of the ensemble quantile classifier consistently estimate the minimal
population loss under suitable general model assumptions. It is also shown that
the ensemble quantile classifier is Bayes optimal under suitable assumptions
with asymmetric Laplace distribution inputs.
| stat |
Social and material vulnerability in the face of seismic hazard: an
analysis of the Italian case | The assessment of the vulnerability of a community endangered by seismic
hazard is of paramount importance for planning a precision policy aimed at the
prevention and reduction of its seismic risk. We aim at measuring the
vulnerability of the Italian municipalities exposed to seismic hazard, by
analyzing the open data offered by the Mappa dei Rischi dei Comuni Italiani
provided by ISTAT, the Italian National Institute of Statistics. Encompassing
the Index of Social and Material Vulnerability already computed by ISTAT, we
also consider as referents of the latent social and material vulnerability of a
community, its demographic dynamics and the age of the building stock where the
community resides. Fusing the analyses of different indicators, within the
context of seismic risk we offer a tentative ranking of the Italian
municipalities in terms of their social and material vulnerability, together
with differential profiles of their dominant fragilities which constitute the
basis for planning precision policies aimed at seismic risk prevention and
reduction.
| stat |
Human mortality at extreme age | We use a combination of extreme value theory, survival analysis and
computer-intensive methods to analyze the mortality of Italian and French
semi-supercentenarians for whom there are validated records. After accounting
for the effects of the sampling frame, there appears to be a constant rate of
mortality beyond age 108 years and no difference between countries and cohorts.
These findings are consistent with previous work based on the International
Database on Longevity and suggest that any physical upper bound for humans is
so large that it is unlikely to be approached. There is no evidence of
differences in survival between women and men after age 108 in the Italian data
and the International Database on Longevity; however survival is lower for men
in the French data.
| stat |
Network Analysis of Orchestral Concert Programming | Orchestral concert programming is a challenging, yet critical task for
expanding audience engagement and is usually driven by qualitative heuristics
and common musical practices. Quantitative analysis of orchestral programming
has been limited, but has become more possible as many orchestras archive their
performance history online. The contribution of this work is to use statistical
network models to quantitatively explore orchestral concert programming,
focusing on which factors determine if two composers are programmed together in
the same concert by the Boston Symphony Orchestra. We find that the type of
composition is the most important covariate in determining which composers are
performed together and the additive and multiplicative effects are logical from
an orchestral programming perspective. These results suggest that a network
analysis is a promising approach for the analysis of concert programming, with
several directions for future extensions.
| stat |
Quantifying and Detecting Individual Level `Always Survivor' Causal
Effects Under `Truncation by Death' and Censoring Through Time | The analysis of causal effects when the outcome of interest is possibly
truncated by death has a long history in statistics and causal inference. The
survivor average causal effect is commonly identified with more assumptions
than those guaranteed by the design of a randomized clinical trial or using
sensitivity analysis. This paper demonstrates that individual level causal
effects in the `always survivor' principal stratum can be identified with no
stronger identification assumptions than randomization. We illustrate the
practical utility of our methods using data from a clinical trial on patients
with prostate cancer. Our methodology is the first and, as of yet, only
proposed procedure that enables detecting causal effects in the presence of
truncation by death using only the assumptions that are guaranteed by design of
the clinical trial. This methodology is applicable to all types of outcomes.
| stat |
Can Bayes Factors "Prove" the Null Hypothesis? | It is possible to obtain a large Bayes Factor (BF) favoring the null
hypothesis when both the null and alternative hypotheses have low likelihoods,
and there are other hypotheses being ignored that are much more strongly
supported by the data. As sample sizes become large it becomes increasingly
probable that a strong BF favouring a point null against a conventional
Bayesian vague alternative co-occurs with a BF favouring various specific
alternatives against the null. For any BF threshold q and sample mean, there is
a value n such that sample sizes larger than n guarantee that although the BF
comparing H0 against a conventional (vague) alternative exceeds q, nevertheless
for some range of hypothetical {\mu}, a BF comparing H0 against {\mu} in that
range falls below 1/q. This paper discusses the conditions under which this
conundrum occurs and investigates methods for resolving it.
| stat |
The Effect of Prior Lipschitz Continuity on the Adversarial Robustness
of Bayesian Neural Networks | It is desirable, and often a necessity, for machine learning models to be
robust against adversarial attacks. This is particularly true for Bayesian
models, as they are well-suited for safety-critical applications, in which
adversarial attacks can have catastrophic outcomes. In this work, we take a
deeper look at the adversarial robustness of Bayesian Neural Networks (BNNs).
In particular, we consider whether the adversarial robustness of a BNN can be
increased by model choices, particularly the Lipschitz continuity induced by
the prior. Conducting in-depth analysis on the case of i.i.d., zero-mean
Gaussian priors and posteriors approximated via mean-field variational
inference, we find evidence that adversarial robustness is indeed sensitive to
the prior variance.
| stat |
Fast calibrated additive quantile regression | We propose a novel framework for fitting additive quantile regression models,
which provides well calibrated inference about the conditional quantiles and
fast automatic estimation of the smoothing parameters, for model structures as
diverse as those usable with distributional GAMs, while maintaining equivalent
numerical efficiency and stability. The proposed methods are at once
statistically rigorous and computationally efficient, because they are based on
the general belief updating framework of Bissiri et al. (2016) to loss based
inference, but compute by adapting the stable fitting methods of Wood et al.
(2016). We show how the pinball loss is statistically suboptimal relative to a
novel smooth generalisation, which also gives access to fast estimation
methods. Further, we provide a novel calibration method for efficiently
selecting the 'learning rate' balancing the loss with the smoothing priors
during inference, thereby obtaining reliable quantile uncertainty estimates.
Our work was motivated by a probabilistic electricity load forecasting
application, used here to demonstrate the proposed approach. The methods
described here are implemented by the qgam R package, available on the
Comprehensive R Archive Network (CRAN).
| stat |
Targeted Principal Components Regression | We propose a principal components regression method based on maximizing a
joint pseudo-likelihood for responses and predictors. Our method uses both
responses and predictors to select linear combinations of the predictors
relevant for the regression, thereby addressing an oft-cited deficiency of
conventional principal components regression. The proposed estimator is shown
to be consistent in a wide range of settings, including ones with non-normal
and dependent observations; conditions on the first and second moments suffice
if the number of predictors ($p$) is fixed and the number of observations ($n$)
tends to infinity and dependence is weak, while stronger distributional
assumptions are needed when $p \to \infty$ with $n$. We obtain the estimator's
asymptotic distribution as the projection of a multivariate normal random
vector onto a tangent cone of the parameter set at the true parameter, and find
the estimator is asymptotically more efficient than competing ones. In
simulations our method is substantially more accurate than conventional
principal components regression and compares favorably to partial least squares
and predictor envelopes. The method is illustrated in a data example with
cross-sectional prediction of stock returns.
| stat |
Moment-Matching Graph-Networks for Causal Inference | In this note we explore a fully unsupervised deep-learning framework for
simulating non-linear structural equation models from observational training
data. The main contribution of this note is an architecture for applying
moment-matching loss functions to the edges of a causal Bayesian graph,
resulting in a generative conditional-moment-matching graph-neural-network.
This framework thus enables automated sampling of latent space conditional
probability distributions for various graphical interventions, and is capable
of generating out-of-sample interventional probabilities that are often
faithful to the ground truth distributions well beyond the range contained in
the training set. These methods could in principle be used in conjunction with
any existing autoencoder that produces a latent space representation containing
causal graph structures.
| stat |
Envelopes for multivariate linear regression with linearly constrained
coefficients | A constrained multivariate linear model is a multivariate linear model with
the columns of its coefficient matrix constrained to lie in a known subspace.
This class of models includes those typically used to study growth curves and
longitudinal data. Envelope methods have been proposed to improve estimation
efficiency in the class of unconstrained multivariate linear models, but have
not yet been developed for constrained models that we develop in this article.
We first compare the standard envelope estimator based on an unconstrained
multivariate model with the standard estimator arising from a constrained
multivariate model in terms of bias and efficiency. Then, to further improve
efficiency, we propose a novel envelope estimator based on a constrained
multivariate model. Novel envelope-based testing methods are also proposed. We
provide support for our proposals by simulations and by studying the classical
dental data and data from the China Health and Nutrition Survey and a study of
probiotic capacity to reduced Salmonella infection.
| stat |
RCRnorm: An integrated system of random-coefficient hierarchical
regression models for normalizing NanoString nCounter data | Formalin-fixed paraffin-embedded (FFPE) samples have great potential for
biomarker discovery, retrospective studies and diagnosis or prognosis of
diseases. Their application, however, is hindered by the unsatisfactory
performance of traditional gene expression profiling techniques on damaged
RNAs. NanoString nCounter platform is well suited for profiling of FFPE samples
and measures gene expression with high sensitivity which may greatly facilitate
realization of scientific and clinical values of FFPE samples. However,
methodological development for normalization, a critical step when analyzing
this type of data, is far behind. Existing methods designed for the platform
use information from different types of internal controls separately and rely
on an overly-simplified assumption that expression of housekeeping genes is
constant across samples for global scaling. Thus, these methods are not
optimized for the nCounter system, not mentioning that they were not developed
for FFPE samples. We construct an integrated system of random-coefficient
hierarchical regression models to capture main patterns and characteristics
observed from NanoString data of FFPE samples and develop a Bayesian approach
to estimate parameters and normalize gene expression across samples. Our
method, labeled RCRnorm, incorporates information from all aspects of the
experimental design and simultaneously removes biases from various sources. It
eliminates the unrealistic assumption on housekeeping genes and offers great
interpretability. Furthermore, it is applicable to freshly frozen or like
samples that can be generally viewed as a reduced case of FFPE samples.
Simulation and applications showed the superior performance of RCRnorm.
| stat |
Causal inference for social network data | We describe semiparametric estimation and inference for causal effects using
observational data from a single social network. Our asymptotic result is the
first to allow for dependence of each observation on a growing number of other
units as sample size increases. While previous methods have generally
implicitly focused on one of two possible sources of dependence among social
network observations, we allow for both dependence due to transmission of
information across network ties, and for dependence due to latent similarities
among nodes sharing ties. We describe estimation and inference for new causal
effects that are specifically of interest in social network settings, such as
interventions on network ties and network structure. Using our methods to
reanalyze the Framingham Heart Study data used in one of the most influential
and controversial causal analyses of social network data, we find that after
accounting for network structure there is no evidence for the causal effects
claimed in the original paper.
| stat |
Nonparametric Regression with Adaptive Truncation via a Convex
Hierarchical Penalty | We consider the problem of non-parametric regression with a potentially large
number of covariates. We propose a convex, penalized estimation framework that
is particularly well-suited for high-dimensional sparse additive models. The
proposed approach combines appealing features of finite basis representation
and smoothing penalties for non-parametric estimation. In particular, in the
case of additive models, a finite basis representation provides a parsimonious
representation for fitted functions but is not adaptive when component
functions posses different levels of complexity. On the other hand, a smoothing
spline type penalty on the component functions is adaptive but does not offer a
parsimonious representation of the estimated function. The proposed approach
simultaneously achieves parsimony and adaptivity in a computationally efficient
framework. We demonstrate these properties through empirical studies on both
real and simulated datasets. We show that our estimator converges at the
minimax rate for functions within a hierarchical class. We further establish
minimax rates for a large class of sparse additive models. The proposed method
is implemented using an efficient algorithm that scales similarly to the Lasso
with the number of covariates and samples size.
| stat |
Filtering the intensity of public concern from social media count data
with jumps | Count time series obtained from online social media data, such as Twitter,
have drawn increasing interest among academics and market analysts over the
past decade. Transforming Web activity records into counts yields time series
with peculiar features, including the coexistence of smooth paths and sudden
jumps, as well as cross-sectional and temporal dependence. Using Twitter posts
about country risks for the United Kingdom and the United States, this paper
proposes an innovative state space model for multivariate count data with
jumps. We use the proposed model to assess the impact of public concerns in
these countries on market systems. To do so, public concerns inferred from
Twitter data are unpacked into country-specific persistent terms, risk social
amplification events, and co-movements of the country series. The identified
components are then used to investigate the existence and magnitude of
country-risk spillovers and social amplification effects on the volatility of
financial markets.
| stat |
Variation Pattern Classification of Functional Data | A new classification method for functional data is developed for the case
where different groups or classes of functions have similar mean functions but
potentially different second moments. The proposed method, second moment-based
functional classifier (SMFC), uses the Hilbert-Schmidt norm to measure the
discrepancy between the second moment structure of the different groups. The
proposed SMFC method is demonstrated to be sensitive to the discrepancy in the
second moment structure and thus produces lower rate of misclassification
compared to competitor methods. One important innovation lies in the dimension
reduction step where the SMFC method data-adaptively determines the basis
functions that account for most of the discrepancy. Consequently, the
misclassification rate is reduced because it removes components of the
functional data that are only weakly discriminatory. In addition, the selected
discriminative basis functions provide insights on the discrepancy between
groups as the basis functions reveal the features of variation pattern that
differentiate groups. Consistency properties are established and, furthermore,
simulation studies and analysis of phoneme and rat brain activity trajectories
empirically demonstrate the advantages of the proposed method.
| stat |
Bayesian Learning of Neural Network Architectures | In this paper we propose a Bayesian method for estimating architectural
parameters of neural networks, namely layer size and network depth. We do this
by learning concrete distributions over these parameters. Our results show that
regular networks with a learnt structure can generalise better on small
datasets, while fully stochastic networks can be more robust to parameter
initialisation. The proposed method relies on standard neural variational
learning and, unlike randomised architecture search, does not require a
retraining of the model, thus keeping the computational overhead at minimum.
| stat |
Modelling intransitivity in pairwise comparisons with application to
baseball data | In most commonly used ranking systems, some level of underlying transitivity
is assumed. If transitivity exists in a system then information about pairwise
comparisons can be translated to other linked pairs. For example, if typically
A beats B and B beats C, this could inform us about the expected outcome
between A and C. We show that in the seminal Bradley-Terry model knowing the
probabilities of A beating B and B beating C completely defines the probability
of A beating C, with these probabilities determined by individual skill levels
of A, B and C. Users of this model tend not to investigate the validity of this
transitive assumption, nor that some skill levels may not be statistically
significantly different from each other; the latter leading to false
conclusions about rankings. We provide a novel extension to the Bradley-Terry
model, which accounts for both of these features: the intransitive
relationships between pairs of objects are dealt with through interaction terms
that are specific to each pair; and by partitioning the $n$ skills into
$A+1\leq n$ distinct clusters, any differences in the objects' skills become
significant, given appropriate $A$. With $n$ competitors there are $n(n-1)/2$
interactions, so even in multiple round robin competitions this gives too many
parameters to efficiently estimate. Therefore we separately cluster the
$n(n-1)/2$ values of intransitivity into $K$ clusters, giving $(A,K)$
estimatable values respectively, typically with $A+K<n$. Using a Bayesian
hierarchical model, $(A,K)$ are treated as unknown, and inference is conducted
via a reversible jump Markov chain Monte Carlo (RJMCMC) algorithm. The model is
shown to have an improved fit out of sample in both simulated data and when
applied to American League baseball data.
| stat |
Group Inverse-Gamma Gamma Shrinkage for Sparse Regression with
Block-Correlated Predictors | Heavy-tailed continuous shrinkage priors, such as the horseshoe prior, are
widely used for sparse estimation problems. However, there is limited work
extending these priors to predictors with grouping structures. Of particular
interest in this article, is regression coefficient estimation where pockets of
high collinearity in the covariate space are contained within known covariate
groupings. To assuage variance inflation due to multicollinearity we propose
the group inverse-gamma gamma (GIGG) prior, a heavy-tailed prior that can
trade-off between local and group shrinkage in a data adaptive fashion. A
special case of the GIGG prior is the group horseshoe prior, whose shrinkage
profile is correlated within-group such that the regression coefficients
marginally have exact horseshoe regularization. We show posterior consistency
for regression coefficients in linear regression models and posterior
concentration results for mean parameters in sparse normal means models. The
full conditional distributions corresponding to GIGG regression can be derived
in closed form, leading to straightforward posterior computation. We show that
GIGG regression results in low mean-squared error across a wide range of
correlation structures and within-group signal densities via simulation. We
apply GIGG regression to data from the National Health and Nutrition
Examination Survey for associating environmental exposures with liver
functionality.
| stat |
Interpretable Random Forests via Rule Extraction | We introduce SIRUS (Stable and Interpretable RUle Set) for regression, a
stable rule learning algorithm which takes the form of a short and simple list
of rules. State-of-the-art learning algorithms are often referred to as "black
boxes" because of the high number of operations involved in their prediction
process. Despite their powerful predictivity, this lack of interpretability may
be highly restrictive for applications with critical decisions at stake. On the
other hand, algorithms with a simple structure-typically decision trees, rule
algorithms, or sparse linear models-are well known for their instability. This
undesirable feature makes the conclusions of the data analysis unreliable and
turns out to be a strong operational limitation. This motivates the design of
SIRUS, which combines a simple structure with a remarkable stable behavior when
data is perturbed. The algorithm is based on random forests, the predictive
accuracy of which is preserved. We demonstrate the efficiency of the method
both empirically (through experiments) and theoretically (with the proof of its
asymptotic stability). Our R/C++ software implementation sirus is available
from CRAN.
| stat |
Actionable Recourse in Linear Classification | Machine learning models are increasingly used to automate decisions that
affect humans - deciding who should receive a loan, a job interview, or a
social service. In such applications, a person should have the ability to
change the decision of a model. When a person is denied a loan by a credit
score, for example, they should be able to alter its input variables in a way
that guarantees approval. Otherwise, they will be denied the loan as long as
the model is deployed. More importantly, they will lack the ability to
influence a decision that affects their livelihood.
In this paper, we frame these issues in terms of recourse, which we define as
the ability of a person to change the decision of a model by altering
actionable input variables (e.g., income vs. age or marital status). We present
integer programming tools to ensure recourse in linear classification problems
without interfering in model development. We demonstrate how our tools can
inform stakeholders through experiments on credit scoring problems. Our results
show that recourse can be significantly affected by standard practices in model
development, and motivate the need to evaluate recourse in practice.
| stat |
Generalized Low-rank plus Sparse Tensor Estimation by Fast Riemannian
Optimization | We investigate a generalized framework to estimate a latent low-rank plus
sparse tensor, where the low-rank tensor often captures the multi-way principal
components and the sparse tensor accounts for potential model
mis-specifications or heterogeneous signals that are unexplainable by the
low-rank part. The framework is flexible covering both linear and non-linear
models, and can easily handle continuous or categorical variables. We propose a
fast algorithm by integrating the Riemannian gradient descent and a novel
gradient pruning procedure. Under suitable conditions, the algorithm converges
linearly and can simultaneously estimate both the low-rank and sparse tensors.
The statistical error bounds of final estimates are established in terms of the
gradient of loss function. The error bounds are generally sharp under specific
statistical models, e.g., the robust tensor PCA and the community detection in
hypergraph networks with outlier vertices. Moreover, our method achieves
non-trivial error bounds for heavy-tailed tensor PCA whenever the noise has a
finite $2+\varepsilon$ moment. We apply our method to analyze the international
trade flow dataset and the statistician hypergraph co-authorship network, both
yielding new and interesting findings.
| stat |
Testing for stationarity of functional time series in the frequency
domain | Interest in functional time series has spiked in the recent past with papers
covering both methodology and applications being published at a much increased
pace. This article contributes to the research in this area by proposing a new
stationarity test for functional time series based on frequency domain methods.
The proposed test statistics is based on joint dimension reduction via
functional principal components analysis across the spectral density operators
at all Fourier frequencies, explicitly allowing for frequency-dependent levels
of truncation to adapt to the dynamics of the underlying functional time
series. The properties of the test are derived both under the null hypothesis
of stationary functional time series and under the smooth alternative of
locally stationary functional time series. The methodology is theoretically
justified through asymptotic results. Evidence from simulation studies and an
application to annual temperature curves suggests that the test works well in
finite samples.
| stat |
Rethinking the Funding Line at the Swiss National Science Foundation:
Bayesian Ranking and Lottery | Funding agencies rely on peer review and expert panels to select the research
deserving funding. Peer review has limitations, including bias against risky
proposals or interdisciplinary research. The inter-rater reliability between
reviewers and panels is low, particularly for proposals near the funding line.
Funding agencies are increasingly acknowledging the role of chance. The Swiss
National Science Foundation (SNSF) introduced a lottery for proposals in the
middle group of good but not excellent proposals. In this article, we introduce
a Bayesian hierarchical model for the evaluation process. To rank the
proposals, we estimate their expected ranks (ER), which incorporates both the
magnitude and uncertainty of the estimated differences between proposals. A
provisional funding line is defined based on ER and budget. The ER and its
credible interval are used to identify proposals with similar quality and
credible intervals that overlap with the funding line. These proposals are
entered into a lottery. We illustrate the approach for two SNSF grant schemes
in career and project funding. We argue that the method could reduce bias in
the evaluation process. R code, data and other materials for this article are
available online.
| stat |
Combining social media and survey data to nowcast migrant stocks in the
United States | Measuring and forecasting migration patterns, and how they change over time,
has important implications for understanding broader population trends, for
designing policy effectively and for allocating resources. However, data on
migration and mobility are often lacking, and those that do exist are not
available in a timely manner. Social media data offer new opportunities to
provide more up-to-date demographic estimates and to complement more
traditional data sources. Facebook, for example, can be thought of as a large
digital census that is regularly updated. However, its users are not
representative of the underlying population. This paper proposes a statistical
framework to combine social media data with traditional survey data to produce
timely `nowcasts' of migrant stocks by state in the United States. The model
incorporates bias adjustment of the Facebook data, and a pooled principal
component time series approach, to account for correlations across age, time
and space. We illustrate the results for migrants from Mexico, India and
Germany, and show that the model outperforms alternatives that rely solely on
either social media or survey data.
| stat |
The experiment is just as important as the likelihood in understanding
the prior: A cautionary note on robust cognitive modelling | Cognitive modelling shares many features with statistical modelling, making
it seem trivial to borrow from the practices of robust Bayesian statistics to
protect the practice of robust cognitive modelling. We take one aspect of
statistical workflow-prior predictive checks-and explore how they might be
applied to a cognitive modelling task. We find that it is not only the
likelihood that is needed to interpret the priors, we also need to incorporate
experiment information as well. This suggests that while cognitive modelling
might borrow from statistical practices, especially workflow, care must be made
to make the adaptions necessary.
| stat |
A better method to enforce monotonic constraints in regression and
classification trees | In this report we present two new ways of enforcing monotone constraints in
regression and classification trees. One yields better results than the current
LightGBM, and has a similar computation time. The other one yields even better
results, but is much slower than the current LightGBM. We also propose a
heuristic that takes into account that greedily splitting a tree by choosing a
monotone split with respect to its immediate gain is far from optimal. Then, we
compare the results with the current implementation of the constraints in the
LightGBM library, using the well known Adult public dataset. Throughout the
report, we mostly focus on the implementation of our methods that we made for
the LightGBM library, even though they are general and could be implemented in
any regression or classification tree. The best method we propose (a smarter
way to split the tree coupled to a penalization of monotone splits)
consistently beats the current implementation of LightGBM. With small or
average trees, the loss reduction can be as high as 1% in the early stages of
training and decreases to around 0.1% at the loss peak for the Adult dataset.
The results would be even better with larger trees. In our experiments, we
didn't do a lot of tuning of the regularization parameters, and we wouldn't be
surprised to see that increasing the performance of our methods on test sets.
| stat |
Deep Learning for Quantile Regression under Right Censoring:
DeepQuantreg | The computational prediction algorithm of neural network, or deep learning,
has drawn much attention recently in statistics as well as in image recognition
and natural language processing. Particularly in statistical application for
censored survival data, the loss function used for optimization has been mainly
based on the partial likelihood from Cox's model and its variations to utilize
existing neural network library such as Keras, which was built upon the open
source library of TensorFlow. This paper presents a novel application of the
neural network to the quantile regression for survival data with right
censoring, which is adjusted by the inverse of the estimated censoring
distribution in the check function. The main purpose of this work is to show
that the deep learning method could be flexible enough to predict nonlinear
patterns more accurately compared to existing quantile regression methods such
as traditional linear quantile regression and nonparametric quantile regression
with total variation regularization, emphasizing practicality of the method for
censored survival data. Simulation studies were performed to generate nonlinear
censored survival data and compare the deep learning method with existing
quantile regression methods in terms of prediction accuracy. The proposed
method is illustrated with two publicly available breast cancer data sets with
gene signatures. The method has been built into a package and is freely
available at \url{https://github.com/yicjia/DeepQuantreg}.
| stat |
Effect of social isolation in dengue cases in the state of Sao Paulo,
Brazil: an analysis during the COVID-19 pandemic | Background: Studies have shown that human mobility is an important factor in
dengue epidemiology. Changes in mobility resulting from COVID-19 pandemic set
up a real-life situation to test this hypothesis. Our objective was to evaluate
the effect of reduced mobility due to this pandemic in the occurrence of dengue
in the state of S\~ao Paulo, Brazil. Method: It is an ecological study of time
series, developed between January and August 2020. We use the number of
confirmed dengue cases and residential mobility, on a daily basis, from
secondary information sources. Mobility was represented by the daily percentage
variation of residential population isolation, obtained from the Google
database. We modeled the relationship between dengue occurrence and social
distancing by negative binomial regression, adjusted for seasonality. We
represent the social distancing dichotomously (isolation versus no isolation)
and consider lag for isolation from the dates of occurrence of dengue. Results:
The risk of dengue decreased around 9.1% (95% CI: 14.2 to 3.7) in the presence
of isolation, considering a delay of 20 days between the degree of isolation
and the dengue first symptoms. Conclusions: We have shown that mobility can
play an important role in the epidemiology of dengue and should be considered
in surveillance and control activities
| stat |
Bayesian Synthetic Likelihood Estimation for Underreported
Non-Stationary Time Series: Covid-19 Incidence in Spain | The problem of dealing with misreported data is very common in a wide range
of contexts for different reasons. The current situation caused by the Covid-19
worldwide pandemic is a clear example, where the data provided by official
sources were not always reliable due to data collection issues and to the high
proportion of asymptomatic cases. In this work, we explore the performance of
Bayesian Synthetic Likelihood to estimate the parameters of a model capable of
dealing with misreported information and to reconstruct the most likely
evolution of the phenomenon. The performance of the proposed methodology is
evaluated through a comprehensive simulation study and illustrated by
reconstructing the weekly Covid-19 incidence in each Spanish Autonomous
Community in 2020.
| stat |
Unlabeled Data Improves Adversarial Robustness | We demonstrate, theoretically and empirically, that adversarial robustness
can significantly benefit from semisupervised learning. Theoretically, we
revisit the simple Gaussian model of Schmidt et al. that shows a sample
complexity gap between standard and robust classification. We prove that
unlabeled data bridges this gap: a simple semisupervised learning procedure
(self-training) achieves high robust accuracy using the same number of labels
required for achieving high standard accuracy. Empirically, we augment CIFAR-10
with 500K unlabeled images sourced from 80 Million Tiny Images and use robust
self-training to outperform state-of-the-art robust accuracies by over 5 points
in (i) $\ell_\infty$ robustness against several strong attacks via adversarial
training and (ii) certified $\ell_2$ and $\ell_\infty$ robustness via
randomized smoothing. On SVHN, adding the dataset's own extra training set with
the labels removed provides gains of 4 to 10 points, within 1 point of the gain
from using the extra labels.
| stat |
Weighted Lasso Estimates for Sparse Logistic Regression: Non-asymptotic
Properties with Measurement Error | When we are interested in high-dimensional system and focus on classification
performance, the $\ell_{1}$-penalized logistic regression is becoming important
and popular. However, the Lasso estimates could be problematic when penalties
of different coefficients are all the same and not related to the data. We
proposed two types of weighted Lasso estimates depending on covariates by the
McDiarmid inequality. Given sample size $n$ and dimension of covariates $p$,
the finite sample behavior of our proposed methods with a diverging number of
predictors is illustrated by non-asymptotic oracle inequalities such as
$\ell_{1}$-estimation error and squared prediction error of the unknown
parameters. We compare the performance of our methods with former weighted
estimates on simulated data, then apply these methods to do real data analysis.
| stat |
Diffusion approximations and control variates for MCMC | A new methodology is presented for the construction of control variates to
reduce the variance of additive functionals of Markov Chain Monte Carlo (MCMC)
samplers. Our control variates are definedthrough the minimization of the
asymptotic variance of the Langevin diffusion over a family of functions, which
can be seen as a quadratic risk minimization procedure. The use of these
control variates is theoretically justified. We show that the asymptotic
variances of some well-known MCMC algorithms, including the Random Walk
Metropolis and the (Metropolis) Unadjusted/Adjusted Langevin Algorithm, are
close to the asymptotic variance of the Langevin diffusion. Several examples of
Bayesian inference problems demonstrate that the corresponding reduction in the
variance is significant.
| stat |
A Text Mining Discovery of Similarities and Dissimilarities Among Sacred
Scriptures | The careful examination of sacred texts gives valuable insights into human
psychology, different ideas regarding the organization of societies as well as
into terms like truth and God. To improve and deepen our understanding of
sacred texts, their comparison, and their separation is crucial. For this
purpose, we use our data set has nine sacred scriptures. This work deals with
the separation of the Quran, the Asian scriptures Tao-Te-Ching, the Buddhism,
the Yogasutras, and the Upanishads as well as the four books from the Bible,
namely the Book of Proverbs, the Book of Ecclesiastes, the Book of
Ecclesiasticus, and the Book of Wisdom. These scriptures are analyzed based on
the natural language processing NLP creating the mathematical representation of
the corpus in terms of frequencies called document term matrix (DTM). After
this analysis, machine learning methods like supervised and unsupervised
learning are applied to perform classification. Here we use the Multinomial
Naive Bayes (MNB), the Super Vector Machine (SVM), the Random Forest (RF), and
the K-nearest Neighbors (KNN). We obtain that among these methods MNB is able
to predict the class of a sacred text with an accuracy of about 85.84 %.
| stat |
A Matching Procedure for Sequential Experiments that Iteratively Learns
which Covariates Improve Power | We propose a dynamic allocation procedure that increases power and efficiency
when measuring an average treatment effect in sequential randomized trials
exploiting some subjects' previous assessed responses. Subjects arrive
iteratively and are either randomized or paired to a previously randomized
subject and administered the alternate treatment. The pairing is made via a
dynamic matching criterion that iteratively learns which specific covariates
are important to the response. We develop estimators for the average treatment
effect as well as an exact test. We illustrate our method's increase in
efficiency and power over other allocation procedures in both simulated
scenarios and a clinical trial dataset. An R package SeqExpMatch for use by
practitioners is available.
| stat |
Making the most of your day: online learning for optimal allocation of
time | We study online learning for optimal allocation when the resource to be
allocated is time. Examples of possible applications include a driver filling a
day with rides, a landlord renting an estate, etc. Following our initial
motivation, a driver receives ride proposals sequentially according to a
Poisson process and can either accept or reject a proposed ride. If she accepts
the proposal, she is busy for the duration of the ride and obtains a reward
that depends on the ride duration. If she rejects it, she remains on hold until
a new ride proposal arrives. We study the regret incurred by the driver first
when she knows her reward function but does not know the distribution of the
ride duration, and then when she does not know her reward function, either.
Faster rates are finally obtained by adding structural assumptions on the
distribution of rides or on the reward function. This natural setting bears
similarities with contextual (one-armed) bandits, but with the crucial
difference that the normalized reward associated to a context depends on the
whole distribution of contexts.
| stat |
Random Spatial Forests | We introduce random spatial forests, a method of bagging regression trees
allowing for spatial correlation. Our main contribution is the development of a
computationally efficient tree building algorithm which selects each split of
the tree adjusting for spatial correlation. We evaluate two different
approaches for estimation of random spatial forests, a pseudo-likelihood
approach combining random forests with kriging and a non-parametric version for
a general class of spatial smoothers. We show improved prediction accuracy of
our method compared to existing two-step approaches combining random forests
and kriging across a range of numerical simulations and demonstrate its
performance on elemental carbon, organic carbon, silicon, and sulfur
measurements across the continental United States from 2009-2010.
| stat |
Profile control chart based on maximum entropy | Monitoring a process over time is so important in manufacturing processes to
reduce the wastage of money and time. The purpose of this article is to monitor
profile coefficients instead of a process mean. In this paper, two methods are
proposed for monitoring the intercept and slope of the simple linear profile,
simultaneously. The first one is linear regression, and another one is the
maximum entropy principle. A simulation study is applied to compare the two
methods in terms of the second type of error and average run length. Finally,
two real examples are presented to demonstrate the ability of the proposed
chart.
| stat |
Penalized linear regression with high-dimensional pairwise screening | In variable selection, most existing screening methods focus on marginal
effects and ignore dependence between covariates. To improve the performance of
selection, we incorporate pairwise effects in covariates for screening and
penalization. We achieve this by studying the asymptotic distribution of the
maximal absolute pairwise sample correlation among independent covariates. The
novelty of the theory is in that the convergence is with respect to the
dimensionality $p$, and is uniform with respect to the sample size $n$.
Moreover, we obtain an upper bound for the maximal pairwise R squared when
regressing the response onto two different covariates. Based on these extreme
value results, we propose a screening procedure to detect covariates pairs that
are potentially correlated and associated with the response. We further combine
the pairwise screening with Sure Independence Screening and develop a new
regularized variable selection procedure. Numerical studies show that our
method is very competitive in terms of both prediction accuracy and variable
selection accuracy.
| stat |
Robust estimation in beta regression via maximum Lq-likelihood | Beta regression models are widely used for modeling continuous data limited
to the unit interval, such as proportions, fractions, and rates. The inference
for the parameters of beta regression models is commonly based on maximum
likelihood estimation. However, it is known to be sensitive to discrepant
observations. In some cases, one atypical data point can lead to severe bias
and erroneous conclusions about the features of interest. In this work, we
develop a robust estimation procedure for beta regression models based on the
maximization of a reparameterized Lq-likelihood. The new estimator offers a
trade-off between robustness and efficiency through a tuning constant. To
select the optimal value of the tuning constant, we propose a data-driven
method which ensures full efficiency in the absence of outliers. We also
improve on an alternative robust estimator by applying our data-driven method
to select its optimum tuning constant. Monte Carlo simulations suggest marked
robustness of the two robust estimators with little loss of efficiency.
Applications to three datasets are presented and discussed. As a by-product of
the proposed methodology, residual diagnostic plots based on robust fits
highlight outliers that would be masked under maximum likelihood estimation.
| stat |
Robust Mixture Modeling using Weighted Complete Estimating Equations | Mixture modeling that takes account of potential heterogeneity in data is
widely adopted for classification and clustering problems. However, it can be
sensitive to outliers especially when the mixture components are Gaussian. In
this paper, we introduce the robust estimating method using the weighted
complete estimating equations for the robust fitting of multivariate mixture
models. The proposed approach is based on a simple modification of the complete
estimating equation given the latent variables of grouping indicators with the
weights that depend on the components of mixture distributions. The weights are
designed in such a way that outliers are downweighted and a simple computing
strategy is facilitated. We develop a simple expectation-estimating-equation
(EEE) algorithm to solve the weighted complete estimating equations. As
examples, the multivariate Gaussian mixture, the mixture of experts, and
multivariate skew normal mixture are considered. In particular, we derive a
novel EEE algorithm for the skew normal mixture which results in the
closed-form expressions for both the E- and EE-steps by slightly extending the
proposed method. The numerical performance of the proposed method is examined
through simulated and real datasets.
| stat |
Gaussian Process Landmarking for Three-Dimensional Geometric
Morphometrics | We demonstrate applications of the Gaussian process-based landmarking
algorithm proposed in [T. Gao, S.Z. Kovalsky, and I. Daubechies, SIAM Journal
on Mathematics of Data Science (2019)] to geometric morphometrics, a branch of
evolutionary biology centered at the analysis and comparisons of anatomical
shapes, and compares the automatically sampled landmarks with the "ground
truth" landmarks manually placed by evolutionary anthropologists; the results
suggest that Gaussian process landmarks perform equally well or better, in
terms of both spatial coverage and downstream statistical analysis. We provide
a detailed exposition of numerical procedures and feature filtering algorithms
for computing high-quality and semantically meaningful diffeomorphisms between
disk-type anatomical surfaces.
| stat |
Probabilistic Canonical Correlation Analysis for Sparse Count Data | Canonical correlation analysis (CCA) is a classical and important
multivariate technique for exploring the relationship between two sets of
continuous variables. CCA has applications in many fields, such as genomics and
neuroimaging. It can extract meaningful features as well as use these features
for subsequent analysis. Although some sparse CCA methods have been developed
to deal with high-dimensional problems, they are designed specifically for
continuous data and do not consider the integer-valued data from
next-generation sequencing platforms that exhibit very low counts for some
important features. We propose a model-based probabilistic approach for
correlation and canonical correlation estimation for two sparse count data sets
(PSCCA). PSCCA demonstrates that correlations and canonical correlations
estimated at the natural parameter level are more appropriate than traditional
estimation methods applied to the raw data. We demonstrate through simulation
studies that PSCCA outperforms other standard correlation approaches and sparse
CCA approaches in estimating the true correlations and canonical correlations
at the natural parameter level. We further apply the PSCCA method to study the
association of miRNA and mRNA expression data sets from a squamous cell lung
cancer study, finding that PSCCA can uncover a large number of strongly
correlated pairs than standard correlation and other sparse CCA approaches.
| stat |
Modelling wind speed with a univariate probability distribution
depending on two baseline functions | Characterizing the wind speed distribution properly is essential for the
satisfactory production of potential energy in wind farms, being the mixture
models usually employed in the description of such data. However, some mixture
models commonly have the undesirable property of non-identifiability. In this
work, we present an alternative distribution which is able to fit the wind
speed data adequately. The new model, called Normal-Weibull-Weibull, is
identifiable and its cumulative distribution function is written as a
composition of two baseline functions. We discuss structural properties of the
class that generates the proposed model, such as the linear representation of
the probability density function, moments and moment generating function. We
perform a Monte Carlo simulation study to investigate the behavior of the
maximum likelihood estimates of the parameters. Finally, we present
applications of the new distribution for modelling wind speed data measured in
five different cities of the Northeastern Region of Brazil.
| stat |
Filtering and Smoothing with Score-Driven Models | We propose a methodology for filtering, smoothing and assessing parameter and
filtering uncertainty in score-driven models. Our technique is based on a
general representation of the Kalman filter and smoother recursions for linear
Gaussian models in terms of the score of the conditional log-likelihood. We
prove that, when data is generated by a nonlinear non-Gaussian state-space
model, the proposed methodology results from a local expansion of the true
filtering density. A formal characterization of the approximation error is
provided. As shown in extensive Monte Carlo analyses, our methodology performs
very similarly to exact simulation-based methods, while remaining
computationally extremely simple. We illustrate empirically the advantages in
employing score-driven models as approximate filters rather than purely
predictive processes.
| stat |
The ordering of future observations from multiple groups | There are many situations where comparison of different groups is of great
interest. Considering the ordering of the efficiency of some treatments is an
example. We present nonparametric predictive inference (NPI) for the ordering
of real-valued future observations from multiple independent groups. The
uncertainty is quantified using NPI lower and upper probabilities for the event
that the next future observations from these groups are ordered in a specific
way. Several applications of these NPI lower and upper probabilities are
explored, including multiple groups inference, diagnostic accuracy and ranked
set sampling.
| stat |
Bayesian inference under small sample size -- A noninformative prior
approach | A Bayesian inference method for problems with small samples and sparse data
is presented in this paper. A general type of prior ($\propto 1/\sigma^{q}$) is
proposed to formulate the Bayesian posterior for inference problems under small
sample size. It is shown that this type of prior can represents a broad range
of priors such as classical noninformative priors and asymptotically locally
invariant priors. It is further shown in this study that such priors can be
derived as the limiting states of Normal-Inverse-Gamma conjugate priors,
allowing for analytical evaluations of Bayesian posteriors and predictors. The
performance of different noninformative priors under small sample size is
compared using the global likelihood. The method of Laplace approximation is
employed to evaluate the global likelihood. A numerical linear regression
problem and a realistic fatigue reliability problem are used to demonstrate the
method and identify the optimal noninformative prior. Results indicate the
predictor using Jeffreys' prior outperforms others. The advantage of the
noninformative Bayesian estimator over the regular least square estimator under
small sample size is shown.
| stat |
Visual High Dimensional Hypothesis Testing | In exploratory data analysis of known classes of high dimensional data, a
central question is how distinct are the classes? The Direction Projection
Permutation (DiProPerm) hypothesis test provides an answer to this that is
directly connected to a visual analysis of the data. In this paper, we propose
an improved DiProPerm test that solves 3 major challenges of the original
version. First, we implement only balanced permutations to increase the test
power for data with strong signals. Second, our mathematical analysis leads to
an adjustment to correct the null behavior of both balanced and the
conventional all permutations. Third, new confidence intervals (reflecting
permutation variation) for test significance are also proposed for comparison
of results across different contexts. This improvement of DiProPerm inference
is illustrated in the context of comparing cancer types in examples from The
Cancer Genome Atlas.
| stat |
Clustering Students and Inferring Skill Set Profiles with Skill
Hierarchies | Cognitive diagnosis models (CDMs) are a popular tool for assessing students'
mastery of sets of skills. Given a set of $K$ skills tested on an assessment,
students are classified into one of $2^K$ latent skill set profiles that
represent whether they have mastered each skill or not. Traditional approaches
to estimating these profiles are computationally intensive and become
infeasible on large datasets. Instead, proxy skill estimates can be generated
from the observed responses and then clustered, and these clusters can be
assigned to different profiles. Building on previous work, we consider how to
optimally perform this clustering when not all $2^K$ profiles are possible,
e.g. because of hierarchical relationships among the skills, and when not all
possible profiles are present in the population. We compare hierarchical
clustering and several k-means variants, including semisupervised clustering
using simulated student responses. The empty k-means algorithm paired with a
novel method for generating starting centers yields the best overall
performance.
| stat |
Active Model Estimation in Markov Decision Processes | We study the problem of efficient exploration in order to learn an accurate
model of an environment, modeled as a Markov decision process (MDP). Efficient
exploration in this problem requires the agent to identify the regions in which
estimating the model is more difficult and then exploit this knowledge to
collect more samples there. In this paper, we formalize this problem, introduce
the first algorithm to learn an $\epsilon$-accurate estimate of the dynamics,
and provide its sample complexity analysis. While this algorithm enjoys strong
guarantees in the large-sample regime, it tends to have a poor performance in
early stages of exploration. To address this issue, we propose an algorithm
that is based on maximum weighted entropy, a heuristic that stems from common
sense and our theoretical analysis. The main idea here is to cover the entire
state-action space with the weight proportional to the noise in the
transitions. Using a number of simple domains with heterogeneous noise in their
transitions, we show that our heuristic-based algorithm outperforms both our
original algorithm and the maximum entropy algorithm in the small sample
regime, while achieving similar asymptotic performance as that of the original
algorithm.
| stat |
Inferring the Type of Phase Transitions Undergone in Epileptic Seizures
Using Random Graph Hidden Markov Models for Percolation in Noisy Dynamic
Networks | In clinical neuroscience, epileptic seizures have been associated with the
sudden emergence of coupled activity across the brain. The resulting functional
networks - in which edges indicate strong enough coupling between brain regions
- are consistent with the notion of percolation, which is a phenomenon in
complex networks corresponding to the sudden emergence of a giant connected
component. Traditionally, work has concentrated on noise-free percolation with
a monotonic process of network growth, but real-world networks are more
complex. We develop a class of random graph hidden Markov models (RG-HMMs) for
characterizing percolation regimes in noisy, dynamically evolving networks in
the presence of edge birth and edge death, as well as noise. This class is used
to understand the type of phase transitions undergone in a seizure, and in
particular, distinguishing between different percolation regimes in epileptic
seizures. We develop a hypothesis testing framework for inferring putative
percolation mechanisms. As a necessary precursor, we present an EM algorithm
for estimating parameters from a sequence of noisy networks only observed at a
longitudinal subsampling of time points. Our results suggest that different
types of percolation can occur in human seizures. The type inferred may suggest
tailored treatment strategies and provide new insights into the fundamental
science of epilepsy.
| stat |
Piecewise Linear Regression via a Difference of Convex Functions | We present a new piecewise linear regression methodology that utilizes
fitting a difference of convex functions (DC functions) to the data. These are
functions $f$ that may be represented as the difference $\phi_1 - \phi_2$ for a
choice of convex functions $\phi_1, \phi_2$. The method proceeds by estimating
piecewise-liner convex functions, in a manner similar to max-affine regression,
whose difference approximates the data. The choice of the function is
regularised by a new seminorm over the class of DC functions that controls the
$\ell_\infty$ Lipschitz constant of the estimate. The resulting methodology can
be efficiently implemented via Quadratic programming even in high dimensions,
and is shown to have close to minimax statistical risk. We empirically validate
the method, showing it to be practically implementable, and to have comparable
performance to existing regression/classification methods on real-world
datasets.
| stat |
Cholesky-based multivariate Gaussian regression | Multivariate Gaussian regression is embedded into a general distributional
regression framework using flexible additive predictors determining all
distributional parameters. While this is relatively straightforward for the
means of the multivariate dependent variable, it is more challenging for the
full covariance matrix {\Sigma} due to two main difficulties: (i) ensuring
positive-definiteness of {\Sigma} and (ii) regularizing the high model
complexity. Both challenges are addressed by adopting a parameterization of
{\Sigma} based on its basic or modified Cholesky decomposition, respectively.
Unlike the decomposition into variances and a correlation matrix, the Cholesky
decomposition guarantees positive-definiteness for any predictor values
regardless of the distributional dimension. Thus, this enables linking all
distributional parameters to flexible predictors without any joint constraints
that would substantially complicate other parameterizations. Moreover, this
approach enables regularization of the flexible additive predictors through
penalized maximum likelihood or Bayesian estimation as for other distributional
regression models. Finally, the Cholesky decomposition allows to reduce the
number of parameters when the components of the multivariate dependent variable
have a natural order (typically time) and a maximum lag can be assumed for the
dependencies among the components.
| stat |
A Bayesian Quest for Finding a Unified Model for Predicting Volleyball
Games | Volleyball is a team sport with unique and specific characteristics. We
introduce a new two level-hierarchical Bayesian model which accounts for theses
volleyball specific characteristics. In the first level, we model the set
outcome with a simple logistic regression model. Conditionally on the winner of
the set, in the second level, we use a truncated negative binomial distribution
for the points earned by the loosing team. An additional Poisson distributed
inflation component is introduced to model the extra points played in the case
that the two teams have point difference less than two points. The number of
points of the winner within each set is deterministically specified by the
winner of the set and the points of the inflation component. The team specific
abilities and the home effect are used as covariates on all layers of the model
(set, point, and extra inflated points). The implementation of the proposed
model on the Italian Superlega 2017/2018 data shows an exceptional
reproducibility of the final league table and a satisfactory predictive
ability.
| stat |
Modal clustering of matrix-variate data | The nonparametric formulation of density-based clustering, known as modal
clustering, draws a correspondence between groups and the attraction domains of
the modes of the density function underlying the data. Its probabilistic
foundation allows for a natural, yet not trivial, generalization of the
approach to the matrix-valued setting, increasingly widespread, for example, in
longitudinal and multivariate spatio-temporal studies. In this work we
introduce nonparametric estimators of matrix-variate distributions based on
kernel methods, and analyze their asymptotic properties. Additionally, we
propose a generalization of the mean-shift procedure for the identification of
the modes of the estimated density. Given the intrinsic high dimensionality of
matrix-variate data, we discuss some locally adaptive solutions to handle the
problem. We test the procedure via extensive simulations, also with respect to
some competitors, and illustrate its performance through two high-dimensional
real data applications.
| stat |
Simulation of an Elevator Group Control Using Generative Adversarial
Networks and Related AI Tools | Testing new, innovative technologies is a crucial task for safety and
acceptance. But how can new systems be tested if no historical real-world data
exist? Simulation provides an answer to this important question. Classical
simulation tools such as event-based simulation are well accepted. But most of
these established simulation models require the specification of many
parameters. Furthermore, simulation runs, e.g., CFD simulations, are very time
consuming. Generative Adversarial Networks (GANs) are powerful tools for
generating new data for a variety of tasks. Currently, their most frequent
application domain is image generation. This article investigates the
applicability of GANs for imitating simulations. We are comparing the
simulation output of a technical system with the output of a GAN. To exemplify
this approach, a well-known multi-car elevator system simulator was chosen. Our
study demonstrates the feasibility of this approach. It also discusses pitfalls
and technical problems that occurred during the implementation. Although we
were able to show that in principle, GANs can be used as substitutes for
expensive simulation runs, we also show that they cannot be used "out of the
box". Fine tuning is needed. We present a proof-of-concept, which can serve as
a starting point for further research.
| stat |
Sequential design of multi-fidelity computer experiments: maximizing the
rate of stepwise uncertainty reduction | This article deals with the sequential design of experiments for
(deterministic or stochastic) multi-fidelity numerical simulators, that is,
simulators that offer control over the accuracy of simulation of the physical
phenomenon or system under study. Very often, accurate simulations correspond
to high computational efforts whereas coarse simulations can be obtained at a
smaller cost. In this setting, simulation results obtained at several levels of
fidelity can be combined in order to estimate quantities of interest (the
optimal value of the output, the probability that the output exceeds a given
threshold...) in an efficient manner. To do so, we propose a new Bayesian
sequential strategy called Maximal Rate of Stepwise Uncertainty Reduction
(MR-SUR), that selects additional simulations to be performed by maximizing the
ratio between the expected reduction of uncertainty and the cost of simulation.
This generic strategy unifies several existing methods, and provides a
principled approach to develop new ones. We assess its performance on several
examples, including a computationally intensive problem of fire safety analysis
where the quantity of interest is the probability of exceeding a tenability
threshold during a building fire.
| stat |
MuyGPs: Scalable Gaussian Process Hyperparameter Estimation Using Local
Cross-Validation | Gaussian processes (GPs) are non-linear probabilistic models popular in many
applications. However, na\"ive GP realizations require quadratic memory to
store the covariance matrix and cubic computation to perform inference or
evaluate the likelihood function. These bottlenecks have driven much investment
in the development of approximate GP alternatives that scale to the large data
sizes common in modern data-driven applications. We present in this manuscript
MuyGPs, a novel efficient GP hyperparameter estimation method. MuyGPs builds
upon prior methods that take advantage of the nearest neighbors structure of
the data, and uses leave-one-out cross-validation to optimize covariance
(kernel) hyperparameters without realizing a possibly expensive likelihood. We
describe our model and methods in detail, and compare our implementations
against the state-of-the-art competitors in a benchmark spatial statistics
problem. We show that our method outperforms all known competitors both in
terms of time-to-solution and the root mean squared error of the predictions.
| stat |
Explaining Machine Learning Models using Entropic Variable Projection | In this paper, we present a new explainability formalism designed to explain
how each input variable of a test set impacts the predictions of machine
learning models. Hence, we propose a group explainability formalism for trained
machine learning decision rules, based on their response to the variability of
the input variables distribution. In order to emphasize the impact of each
input variable, this formalism uses an information theory framework that
quantifies the influence of all input-output observations based on entropic
projections. This is thus the first unified and model agnostic formalism
enabling data scientists to interpret the dependence between the input
variables, their impact on the prediction errors, and their influence on the
output predictions. Convergence rates of the entropic projections are provided
in the large sample case. Most importantly, we prove that computing an
explanation in our framework has a low algorithmic complexity, making it
scalable to real-life large datasets. We illustrate our strategy by explaining
complex decision rules learned by using XGBoost, Random Forest or Deep Neural
Network classifiers on various datasets such as Adult income, MNIST and CelebA.
We finally make clear its differences with the explainability strategies
\textit{LIME} and \textit{SHAP}, that are based on single observations. Results
can be reproduced by using the freely distributed Python toolbox
https://gems-ai.com}.
| stat |
Distributed, partially collapsed MCMC for Bayesian Nonparametrics | Bayesian nonparametric (BNP) models provide elegant methods for discovering
underlying latent features within a data set, but inference in such models can
be slow. We exploit the fact that completely random measures, which commonly
used models like the Dirichlet process and the beta-Bernoulli process can be
expressed as, are decomposable into independent sub-measures. We use this
decomposition to partition the latent measure into a finite measure containing
only instantiated components, and an infinite measure containing all other
components. We then select different inference algorithms for the two
components: uncollapsed samplers mix well on the finite measure, while
collapsed samplers mix well on the infinite, sparsely occupied tail. The
resulting hybrid algorithm can be applied to a wide class of models, and can be
easily distributed to allow scalable inference without sacrificing asymptotic
convergence guarantees.
| stat |
Matching on Generalized Propensity Scores with Continuous Exposures | The robustness of statistical methods and interpretability of statistical
analysis from observational studies are central concerns in many applied
fields, and robust causal inference methods have the potential to mitigate
these concerns. When estimating the causal effects of continuous exposure in
observational studies, generalized propensity scores (GPS) have been used to
adjust for confounding bias. Existing GPS methods, either relying on weighting
or regression, have certain limitations: a) they require a correctly specified
outcome model; b) they are sensitive to extreme values of the estimated GPS; c)
assessing covariate balance when using these approaches is not straightforward.
Matching, a class of popular causal inference methods with binary treatments,
has not been extended to the continuous exposure setting, disregarding its many
attractive features on method robustness and interpretability. In this paper,
we propose an innovative approach for GPS caliper matching in settings with
continuous exposures. We first introduce an assumption of identifiability,
called local weak unconfoundedness, that is less stringent than what is
currently proposed in the literature. Under this assumption and mild smoothness
conditions, we provide theoretical guarantees that our proposed matching
estimators attain consistency and asymptotic normality. In simulations, our
proposed matching estimator outperforms existing methods under settings of
model misspecification and/or in the presence of extreme values of the
estimated GPS in terms of bias reduction, root mean squared error, and overall
achieves excellent covariate balance. We utilize the largest-to-date Medicare
claims data for the entire US from 2000 to 2016 to construct a continuous
causal exposure-response curve for long-term exposure to fine particles
($PM_{2.5}$) on mortality.
| stat |
Workflow Techniques for the Robust Use of Bayes Factors | Inferences about hypotheses are ubiquitous in the cognitive sciences. Bayes
factors provide one general way to compare different hypotheses by their
compatibility with the observed data. Those quantifications can then also be
used to choose between hypotheses. While Bayes factors provide an immediate
approach to hypothesis testing, they are highly sensitive to details of the
data/model assumptions. Moreover it's not clear how straightforwardly this
approach can be implemented in practice, and in particular how sensitive it is
to the details of the computational implementation. Here, we investigate these
questions for Bayes factor analyses in the cognitive sciences. We explain the
statistics underlying Bayes factors as a tool for Bayesian inferences and
discuss that utility functions are needed for principled decisions on
hypotheses. Next, we study how Bayes factors misbehave under different
conditions. This includes a study of errors in the estimation of Bayes factors.
Importantly, it is unknown whether Bayes factor estimates based on bridge
sampling are unbiased for complex analyses. We are the first to use
simulation-based calibration as a tool to test the accuracy of Bayes factor
estimates. Moreover, we study how stable Bayes factors are against different
MCMC draws. We moreover study how Bayes factors depend on variation in the
data. We also look at variability of decisions based on Bayes factors and how
to optimize decisions using a utility function. We outline a Bayes factor
workflow that researchers can use to study whether Bayes factors are robust for
their individual analysis, and we illustrate this workflow using an example
from the cognitive sciences. We hope that this study will provide a workflow to
test the strengths and limitations of Bayes factors as a way to quantify
evidence in support of scientific hypotheses. Reproducible code is available
from https://osf.io/y354c/.
| stat |
Controlling False Discovery Rates Using Null Bootstrapping | We consider controlling the false discovery rate for many tests with unknown
correlation structure. Given a large number of hypotheses, false and missing
discoveries can plague an analysis. While many procedures have been proposed to
control false discovery, they either assume independent hypotheses or lack
statistical power. We propose a novel method for false discovery control using
null bootstrapping. By bootstrapping from the correlated null, we achieve
superior statistical power to existing methods and prove that the false
discovery rate is controlled. Simulated examples illustrate the efficacy of our
method over existing methods. We apply our proposed methodology to financial
asset pricing, where the goal is to determine which "factors" lead to excess
returns out of a large number of potential factors.
| stat |
Spatiotemporal Multi-Resolution Approximations for Analyzing Global
Environmental Data | Technological developments and open data policies have made large, global
environmental datasets accessible to everyone. For analysing such datasets,
including spatiotemporal correlations using traditional models based on
Gaussian processes does not scale with data volume and requires strong
assumptions about stationarity, separability, and distance measures of
covariance functions that are often unrealistic for global data. Only very few
modeling approaches suitably model spatiotemporal correlations while addressing
both computational scalability as well as flexible covariance models. In this
paper, we provide an extension to the multi-resolution approximation (MRA)
approach for spatiotemporal modeling of global datasets. MRA has been shown to
be computationally scalable in distributed computing environments and allows
for integrating arbitrary user-defined covariance functions. Our extension adds
a spatiotemporal partitioning, and fitting of complex covariance models
including nonstationarity with kernel convolutions and spherical distances. We
evaluate the effect of the MRA parameters on estimation and spatiotemporal
prediction using simulated data, where computation times reduced around two
orders of magnitude with an increase of the root-mean-square prediction error
of around five percent. This allows for trading off computation times against
prediction errors, and we derive a practical strategy for selecting the MRA
parameters. We demonstrate how the approach can be practically used for
analyzing daily sea surface temperature and precipitation data on global scale
and compare models with different complexities in the covariance function.
| stat |
Diverse personalized recommendations with uncertainty from implicit
preference data with the Bayesian Mallows Model | Clicking data, which exists in abundance and contains objective user
preference information, is widely used to produce personalized recommendations
in web-based applications. Current popular recommendation algorithms, typically
based on matrix factorizations, often have high accuracy and achieve good
clickthrough rates. However, diversity of the recommended items, which can
greatly enhance user experiences, is often overlooked. Moreover, most
algorithms do not produce interpretable uncertainty quantifications of the
recommendations. In this work, we propose the Bayesian Mallows for Clicking
Data (BMCD) method, which augments clicking data into compatible full ranking
vectors by enforcing all the clicked items to be top-ranked. User preferences
are learned using a Mallows ranking model. Bayesian inference leads to
interpretable uncertainties of each individual recommendation, and we also
propose a method to make personalized recommendations based on such
uncertainties. With a simulation study and a real life data example, we
demonstrate that compared to state-of-the-art matrix factorization, BMCD makes
personalized recommendations with similar accuracy, while achieving much higher
level of diversity, and producing interpretable and actionable uncertainty
estimation.
| stat |
A maximum principle argument for the uniform convergence of graph
Laplacian regressors | This paper investigates the use of methods from partial differential
equations and the Calculus of variations to study learning problems that are
regularized using graph Laplacians. Graph Laplacians are a powerful, flexible
method for capturing local and global geometry in many classes of learning
problems, and the techniques developed in this paper help to broaden the
methodology of studying such problems. In particular, we develop the use of
maximum principle arguments to establish asymptotic consistency guarantees
within the context of noise corrupted, non-parametric regression with samples
living on an unknown manifold embedded in $\mathbb{R}^d$. The maximum principle
arguments provide a new technical tool which informs parameter selection by
giving concrete error estimates in terms of various regularization parameters.
A review of learning algorithms which utilize graph Laplacians, as well as
previous developments in the use of differential equation and variational
techniques to study those algorithms, is given. In addition, new connections
are drawn between Laplacian methods and other machine learning techniques, such
as kernel regression and k-nearest neighbor methods.
| stat |
Syntax-Infused Variational Autoencoder for Text Generation | We present a syntax-infused variational autoencoder (SIVAE), that integrates
sentences with their syntactic trees to improve the grammar of generated
sentences. Distinct from existing VAE-based text generative models, SIVAE
contains two separate latent spaces, for sentences and syntactic trees. The
evidence lower bound objective is redesigned correspondingly, by optimizing a
joint distribution that accommodates two encoders and two decoders. SIVAE works
with long short-term memory architectures to simultaneously generate sentences
and syntactic trees. Two versions of SIVAE are proposed: one captures the
dependencies between the latent variables through a conditional prior network,
and the other treats the latent variables independently such that
syntactically-controlled sentence generation can be performed. Experimental
results demonstrate the generative superiority of SIVAE on both reconstruction
and targeted syntactic evaluations. Finally, we show that the proposed models
can be used for unsupervised paraphrasing given different syntactic tree
templates.
| stat |
Fast Covariance Estimation for Multivariate Sparse Functional Data | Covariance estimation is essential yet underdeveloped for analyzing
multivariate functional data. We propose a fast covariance estimation method
for multivariate sparse functional data using bivariate penalized splines. The
tensor-product B-spline formulation of the proposed method enables a simple
spectral decomposition of the associated covariance operator and explicit
expressions of the resulting eigenfunctions as linear combinations of B-spline
bases, thereby dramatically facilitating subsequent principal component
analysis. We derive a fast algorithm for selecting the smoothing parameters in
covariance smoothing using leave-one-subject-out cross-validation. The method
is evaluated with extensive numerical studies and applied to an Alzheimer's
disease study with multiple longitudinal outcomes.
| stat |
Iterative Alpha Expansion for estimating gradient-sparse signals from
linear measurements | We consider estimating a piecewise-constant image, or a gradient-sparse
signal on a general graph, from noisy linear measurements. We propose and study
an iterative algorithm to minimize a penalized least-squares objective, with a
penalty given by the "l_0-norm" of the signal's discrete graph gradient. The
method proceeds by approximate proximal descent, applying the alpha-expansion
procedure to minimize a proximal gradient in each iteration, and using a
geometric decay of the penalty parameter across iterations. Under a
cut-restricted isometry property for the measurement design, we prove global
recovery guarantees for the estimated signal. For standard Gaussian designs,
the required number of measurements is independent of the graph structure, and
improves upon worst-case guarantees for total-variation (TV) compressed sensing
on the 1-D and 2-D lattice graphs by polynomial and logarithmic factors,
respectively. The method empirically yields lower mean-squared recovery error
compared with TV regularization in regimes of moderate undersampling and
moderate to high signal-to-noise, for several examples of changepoint signals
and gradient-sparse phantom images.
| stat |
Data Fusion for Joining Income and Consumption Information Using
Different Donor-Recipient Distance Metrics | Data fusion describes the method of combining data from (at least) two
initially independent data sources to allow for joint analysis of variables
which are not jointly observed. The fundamental idea is to base inference on
identifying assumptions, and on common variables which provide information that
is jointly observed in all the data sources. A popular class of methods dealing
with this particular missing-data problem is based on nearest neighbour
matching. However, exact matches become unlikely with increasing common
information, and the specification of the distance function can influence the
results of the data fusion. In this paper we compare two different approaches
of nearest neighbour hot deck matching: One, Random Hot Deck, is a variant of
the covariate-based matching methods which was proposed by Eurostat, and can be
considered as a 'classical' statistical matching method, whereas the
alternative approach is based on Predictive Mean Matching. We discuss results
from a simulation study to investigate benefits and potential drawbacks of both
variants, and our findings suggest that Predictive Mean Matching tends to
outperform Random Hot Deck.
| stat |
A Unified Stochastic Gradient Approach to Designing Bayesian-Optimal
Experiments | We introduce a fully stochastic gradient based approach to Bayesian optimal
experimental design (BOED). Our approach utilizes variational lower bounds on
the expected information gain (EIG) of an experiment that can be simultaneously
optimized with respect to both the variational and design parameters. This
allows the design process to be carried out through a single unified stochastic
gradient ascent procedure, in contrast to existing approaches that typically
construct a pointwise EIG estimator, before passing this estimator to a
separate optimizer. We provide a number of different variational objectives
including the novel adaptive contrastive estimation (ACE) bound. Finally, we
show that our gradient-based approaches are able to provide effective design
optimization in substantially higher dimensional settings than existing
approaches.
| stat |
Quasi-stationary Monte Carlo and the ScaLE Algorithm | This paper introduces a class of Monte Carlo algorithms which are based upon
the simulation of a Markov process whose quasi-stationary distribution
coincides with a distribution of interest. This differs fundamentally from,
say, current Markov chain Monte Carlo methods which simulate a Markov chain
whose stationary distribution is the target. We show how to approximate
distributions of interest by carefully combining sequential Monte Carlo methods
with methodology for the exact simulation of diffusions. The methodology
introduced here is particularly promising in that it is applicable to the same
class of problems as gradient based Markov chain Monte Carlo algorithms but
entirely circumvents the need to conduct Metropolis-Hastings type accept/reject
steps whilst retaining exactness: the paper gives theoretical guarantees
ensuring the algorithm has the correct limiting target distribution.
Furthermore, this methodology is highly amenable to big data problems. By
employing a modification to existing na{\"\i}ve sub-sampling and control
variate techniques it is possible to obtain an algorithm which is still exact
but has sub-linear iterative cost as a function of data size.
| stat |
A heavy-tailed and overdispersed collective risk model | Insurance data can be asymmetric with heavy tails, causing inadequate
adjustments of the usually applied models. To deal with this issue,
hierarchical models for collective risk with heavy-tails of the claims
distributions that take also into account overdispersion of the number of
claims are proposed. In particular, the distribution of the logarithm of the
aggregate value of claims is assumed to follow a Student-t distribution.
Additionally, to incorporate possible overdispersion, the number of claims is
modeled as having a negative binomial distribution. Bayesian decision theory is
invoked to calculate the fair premium based on the modified absolute deviation
utility. An application to a health insurance dataset is presented together
with some diagnostic measures to identify excess variability. The variability
measures are analyzed using the marginal posterior predictive distribution of
the premiums according to some competitive models. Finally, a simulation study
is carried out to assess the predictive capability of the model and the
adequacy of the Bayesian estimation procedure.
Keywords: Continuous ranked probability score (CRPS); decision theory;
insurance data; marginal posterior predictive; tail value at risk; value at
risk.
| stat |
Partial Identification of Answer Reviewing Effects in Multiple-Choice
Exams | Does reviewing previous answers during multiple-choice exams help examinees
increase their final score? This article formalizes the question using a
rigorous causal framework, the potential outcomes framework. Viewing examinees'
reviewing status as a treatment and their final score as an outcome, the
article first explains the challenges of identifying the causal effect of
answer reviewing in regular exam-taking settings. In addition to the
incapability of randomizing the treatment selection (reviewing status) and the
lack of other information to make this selection process ignorable, the
treatment variable itself is not fully known to researchers. Looking at
examinees' answer sheet data, it is unclear whether an examinee who did not
change his or her answer on a specific item reviewed it but retained the
initial answer (treatment condition) or chose not to review it (control
condition). Despite such challenges, however, the article develops partial
identification strategies and shows that the sign of the answer reviewing
effect can be reasonably inferred. By analyzing a statewide math assessment
data set, the article finds that reviewing initial answers is generally
beneficial for examinees.
| stat |
Addressing Delayed Feedback for Continuous Training with Neural Networks
in CTR prediction | One of the challenges in display advertising is that the distribution of
features and click through rate (CTR) can exhibit large shifts over time due to
seasonality, changes to ad campaigns and other factors. The predominant
strategy to keep up with these shifts is to train predictive models
continuously, on fresh data, in order to prevent them from becoming stale.
However, in many ad systems positive labels are only observed after a possibly
long and random delay. These delayed labels pose a challenge to data freshness
in continuous training: fresh data may not have complete label information at
the time they are ingested by the training algorithm. Naive strategies which
consider any data point a negative example until a positive label becomes
available tend to underestimate CTR, resulting in inferior user experience and
suboptimal performance for advertisers. The focus of this paper is to identify
the best combination of loss functions and models that enable large-scale
learning from a continuous stream of data in the presence of delayed labels. In
this work, we compare 5 different loss functions, 3 of them applied to this
problem for the first time. We benchmark their performance in offline settings
on both public and proprietary datasets in conjunction with shallow and deep
model architectures. We also discuss the engineering cost associated with
implementing each loss function in a production environment. Finally, we
carried out online experiments with the top performing methods, in order to
validate their performance in a continuous training scheme. While training on
668 million in-house data points offline, our proposed methods outperform
previous state-of-the-art by 3% relative cross entropy (RCE). During online
experiments, we observed 55% gain in revenue per thousand requests (RPMq)
against naive log loss.
| stat |
When and where: estimating the date and location of introduction for
exotic pests and pathogens | A fundamental question during the outbreak of a novel disease or invasion of
an exotic pest is: At what location and date was it first introduced? With this
information, future introductions can be anticipated and perhaps avoided. Point
process models are commonly used for mapping species distribution and disease
occurrence. If the time and location of introductions were known, then point
process models could be used to map and understand the factors that influence
introductions; however, rarely is the process of introduction directly
observed. We propose embedding a point process within hierarchical Bayesian
models commonly used to understand the spatio-temporal dynamics of invasion.
Including a point process within a hierarchical Bayesian model enables
inference regarding the location and date of introduction from indirect
observation of the process such as species or disease occurrence records. We
illustrate our approach using disease surveillance data collected to monitor
white-nose syndrome, which is a fungal disease that threatens many North
American species of bats. We use our model and surveillance data to estimate
the location and date that the pathogen was introduced into the United States.
Finally, we compare forecasts from our model to forecasts obtained from
state-of-the-art regression-based statistical and machine learning methods. Our
results show that the pathogen causing white-nose syndrome was most likely
introduced into the United States 4 years prior to the first detection, but
there is a moderate level of uncertainty in this estimate. The location of
introduction could be up to 510 km east of the location of first discovery, but
our results indicate that there is a relatively high probability the location
of first detection could be the location of introduction.
| stat |
Online Passive-Aggressive Total-Error-Rate Minimization | We provide a new online learning algorithm which utilizes online
passive-aggressive learning (PA) and total-error-rate minimization (TER) for
binary classification. The PA learning establishes not only large margin
training but also the capacity to handle non-separable data. The TER learning
on the other hand minimizes an approximated classification error based
objective function. We propose an online PATER algorithm which combines those
useful properties. In addition, we also present a weighted PATER algorithm to
improve the ability to cope with data imbalance problems. Experimental results
demonstrate that the proposed PATER algorithms achieves better performances in
terms of efficiency and effectiveness than the existing state-of-the-art online
learning algorithms in real-world data sets.
| stat |
Normalizing Flows Across Dimensions | Real-world data with underlying structure, such as pictures of faces, are
hypothesized to lie on a low-dimensional manifold. This manifold hypothesis has
motivated state-of-the-art generative algorithms that learn low-dimensional
data representations. Unfortunately, a popular generative model, normalizing
flows, cannot take advantage of this. Normalizing flows are based on successive
variable transformations that are, by design, incapable of learning
lower-dimensional representations. In this paper we introduce noisy injective
flows (NIF), a generalization of normalizing flows that can go across
dimensions. NIF explicitly map the latent space to a learnable manifold in a
high-dimensional data space using injective transformations. We further employ
an additive noise model to account for deviations from the manifold and
identify a stochastic inverse of the generative process. Empirically, we
demonstrate that a simple application of our method to existing flow
architectures can significantly improve sample quality and yield separable data
embeddings.
| stat |
The Anisotropic Noise in Stochastic Gradient Descent: Its Behavior of
Escaping from Sharp Minima and Regularization Effects | Understanding the behavior of stochastic gradient descent (SGD) in the
context of deep neural networks has raised lots of concerns recently. Along
this line, we study a general form of gradient based optimization dynamics with
unbiased noise, which unifies SGD and standard Langevin dynamics. Through
investigating this general optimization dynamics, we analyze the behavior of
SGD on escaping from minima and its regularization effects. A novel indicator
is derived to characterize the efficiency of escaping from minima through
measuring the alignment of noise covariance and the curvature of loss function.
Based on this indicator, two conditions are established to show which type of
noise structure is superior to isotropic noise in term of escaping efficiency.
We further show that the anisotropic noise in SGD satisfies the two conditions,
and thus helps to escape from sharp and poor minima effectively, towards more
stable and flat minima that typically generalize well. We systematically design
various experiments to verify the benefits of the anisotropic noise, compared
with full gradient descent plus isotropic diffusion (i.e. Langevin dynamics).
| stat |
A Probabilistic framework for Quantum Clustering | Quantum Clustering is a powerful method to detect clusters in data with mixed
density. However, it is very sensitive to a length parameter that is inherent
to the Schr\"odinger equation. In addition, linking data points into clusters
requires local estimates of covariance that are also controlled by length
parameters. This raises the question of how to adjust the control parameters of
the Schr\"odinger equation for optimal clustering. We propose a probabilistic
framework that provides an objective function for the goodness-of-fit to the
data, enabling the control parameters to be optimised within a Bayesian
framework. This naturally yields probabilities of cluster membership and data
partitions with specific numbers of clusters. The proposed framework is tested
on real and synthetic data sets, assessing its validity by measuring
concordance with known data structure by means of the Jaccard score (JS). This
work also proposes an objective way to measure performance in unsupervised
learning that correlates very well with JS.
| stat |
Stochastic Normalizing Flows | The sampling of probability distributions specified up to a normalization
constant is an important problem in both machine learning and statistical
mechanics. While classical stochastic sampling methods such as Markov Chain
Monte Carlo (MCMC) or Langevin Dynamics (LD) can suffer from slow mixing times
there is a growing interest in using normalizing flows in order to learn the
transformation of a simple prior distribution to the given target distribution.
Here we propose a generalized and combined approach to sample target densities:
Stochastic Normalizing Flows (SNF) -- an arbitrary sequence of deterministic
invertible functions and stochastic sampling blocks. We show that stochasticity
overcomes expressivity limitations of normalizing flows resulting from the
invertibility constraint, whereas trainable transformations between sampling
steps improve efficiency of pure MCMC/LD along the flow. By invoking ideas from
non-equilibrium statistical mechanics we derive an efficient training procedure
by which both the sampler's and the flow's parameters can be optimized
end-to-end, and by which we can compute exact importance weights without having
to marginalize out the randomness of the stochastic blocks. We illustrate the
representational power, sampling efficiency and asymptotic correctness of SNFs
on several benchmarks including applications to sampling molecular systems in
equilibrium.
| stat |
Familywise Error Rate Control by Interactive Unmasking | We propose a method for multiple hypothesis testing with familywise error
rate (FWER) control, called the i-FWER test. Most testing methods are
predefined algorithms that do not allow modifications after observing the data.
However, in practice, analysts tend to choose a promising algorithm after
observing the data; unfortunately, this violates the validity of the
conclusion. The i-FWER test allows much flexibility: a human (or a computer
program acting on the human's behalf) may adaptively guide the algorithm in a
data-dependent manner. We prove that our test controls FWER if the analysts
adhere to a particular protocol of "masking" and "unmasking". We demonstrate
via numerical experiments the power of our test under structured non-nulls, and
then explore new forms of masking.
| stat |
A copula-based measure for quantifying asymmetry in dependence and
associations | Asymmetry is an inherent property of bivariate associations and therefore
must not be ignored. The currently applicable dependence measures mask the
potential asymmetry of the underlying dependence structure by implicitly
assuming that quantity Y is equally dependent on quantity X, and vice versa,
which is generally not true. We introduce the copula-based dependence measure
qad that quantifies asymmetry. Specifically, qad is applicable in general
situations, is sensitive to noise in data, detects asymmetry in dependence and
reliably quantifies the information gain/predictability of quantity Y given
knowledge of quantity X, and vice versa. Using real-world data sets, we
demonstrate the relevance of asymmetry in associations. Asymmetry in dependence
is a novel category of information that provides substantial information gain
in analyses of bivariate associations.
| stat |
Disjunct Support Spike and Slab Priors for Variable Selection in
Regression under Quasi-sparseness | Sparseness of the regression coefficient vector is often a desirable
property, since, among other benefits, sparseness improves interpretability. In
practice, many true regression coefficients might be negligibly small, but
non-zero, which we refer to as quasi-sparseness. Spike-and-slab priors as
introduced in (Chipman et al., 2001) can be tuned to ignore very small
regression coefficients, and, as a consequence provide a trade-off between
prediction accuracy and interpretability. However, spike-and-slab priors with
full support lead to inconsistent Bayes factors, in the sense that the Bayes
factors of any two models are bounded in probability. This is clearly an
undesirable property for Bayesian hypotheses testing, where we wish that
increasing sample sizes lead to increasing Bayes factors favoring the true
model. The moment matching priors as in (Johnson and Rossell, 2012) can resolve
this issue, but are unsuitable for the quasi-sparse setting due to their full
support outside the exact value 0. As a remedy, we suggest disjunct support
spike and slab priors, for which we prove consistent Bayes factors in the
quasi-sparse setting, and show experimentally fast growing Bayes factors
favoring the true model. Several experiments on simulated and real data confirm
the usefulness of our proposed method to identify models with high effect size,
while leading to better control over false positives than hard-thresholding.
| stat |
Sequential Drift Detection in Deep Learning Classifiers | We utilize neural network embeddings to detect data drift by formulating the
drift detection within an appropriate sequential decision framework. This
enables control of the false alarm rate although the statistical tests are
repeatedly applied. Since change detection algorithms naturally face a tradeoff
between avoiding false alarms and quick correct detection, we introduce a loss
function which evaluates an algorithm's ability to balance these two concerns,
and we use it in a series of experiments.
| stat |
Robust Adversarial Learning via Sparsifying Front Ends | It is by now well-known that small adversarial perturbations can induce
classification errors in deep neural networks. In this paper, we take a
bottom-up signal processing perspective to this problem and show that a
systematic exploitation of sparsity in natural data is a promising tool for
defense. For linear classifiers, we show that a sparsifying front end is
provably effective against $\ell_{\infty}$-bounded attacks, reducing output
distortion due to the attack by a factor of roughly $K/N$ where $N$ is the data
dimension and $K$ is the sparsity level. We then extend this concept to deep
networks, showing that a "locally linear" model can be used to develop a
theoretical foundation for crafting attacks and defenses. We also devise
attacks based on the locally linear model that outperform the well-known FGSM
attack. We supplement our theoretical results with experiments on the MNIST
handwritten digit database, showing the efficacy of the proposed sparsity-based
defense schemes.
| stat |
On Generalization of Adaptive Methods for Over-parameterized Linear
Regression | Over-parameterization and adaptive methods have played a crucial role in the
success of deep learning in the last decade. The widespread use of
over-parameterization has forced us to rethink generalization by bringing forth
new phenomena, such as implicit regularization of optimization algorithms and
double descent with training progression. A series of recent works have started
to shed light on these areas in the quest to understand -- why do neural
networks generalize well? The setting of over-parameterized linear regression
has provided key insights into understanding this mysterious behavior of neural
networks.
In this paper, we aim to characterize the performance of adaptive methods in
the over-parameterized linear regression setting. First, we focus on two
sub-classes of adaptive methods depending on their generalization performance.
For the first class of adaptive methods, the parameter vector remains in the
span of the data and converges to the minimum norm solution like gradient
descent (GD). On the other hand, for the second class of adaptive methods, the
gradient rotation caused by the pre-conditioner matrix results in an in-span
component of the parameter vector that converges to the minimum norm solution
and the out-of-span component that saturates. Our experiments on
over-parameterized linear regression and deep neural networks support this
theory.
| stat |
Closed Loop Neural-Symbolic Learning via Integrating Neural Perception,
Grammar Parsing, and Symbolic Reasoning | The goal of neural-symbolic computation is to integrate the connectionist and
symbolist paradigms. Prior methods learn the neural-symbolic models using
reinforcement learning (RL) approaches, which ignore the error propagation in
the symbolic reasoning module and thus converge slowly with sparse rewards. In
this paper, we address these issues and close the loop of neural-symbolic
learning by (1) introducing the \textbf{grammar} model as a \textit{symbolic
prior} to bridge neural perception and symbolic reasoning, and (2) proposing a
novel \textbf{back-search} algorithm which mimics the top-down human-like
learning procedure to propagate the error through the symbolic reasoning module
efficiently. We further interpret the proposed learning framework as maximum
likelihood estimation using Markov chain Monte Carlo sampling and the
back-search algorithm as a Metropolis-Hastings sampler. The experiments are
conducted on two weakly-supervised neural-symbolic tasks: (1) handwritten
formula recognition on the newly introduced HWF dataset; (2) visual question
answering on the CLEVR dataset. The results show that our approach
significantly outperforms the RL methods in terms of performance, converging
speed, and data efficiency. Our code and data are released at
\url{https://liqing-ustc.github.io/NGS}.
| stat |
Deductive semiparametric estimation in Double-Sampling Designs with
application to PEPFAR | Non-ignorable dropout is common in studies with long follow-up time, and it
can bias study results unless handled carefully. A double-sampling design
allocates additional resources to pursue a subsample of the dropouts and find
out their outcomes, which can address potential biases due to non-ignorable
dropout. It is desirable to construct semiparametric estimators for the
double-sampling design because of their robustness properties. However,
obtaining such semiparametric estimators remains a challenge due to the
requirement of the analytic form of the efficient influence function (EIF), the
derivation of which can be ad hoc and difficult for the double-sampling design.
Recent work has shown how the derivation of EIF can be made deductive and
computerizable using the functional derivative representation of the EIF in
nonparametric models. This approach, however, requires deriving the mixture of
a continuous distribution and a point mass, which can itself be challenging for
complicated problems such as the double-sampling design. We propose
semiparametric estimators for the survival probability in double-sampling
designs by generalizing the deductive and computerizable estimation approach.
In particular, we propose to build the semiparametric estimators based on a
discretized support structure, which approximates the possibly continuous
observed data distribution and circumvents the derivation of the mixture
distribution. Our approach is deductive in the sense that it is expected to
produce semiparametric locally efficient estimators within finite steps without
knowledge of the EIF. We apply the proposed estimators to estimating the
mortality rate in a double-sampling design component of the President's
Emergency Plan for AIDS Relief (PEPFAR) program. We evaluate the impact of
double-sampling selection criteria on the mortality rate estimates.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.