title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Enhancing statistical inference in psychological research via
prospective and retrospective design analysis | In the past two decades, psychological science has experienced an
unprecedented replicability crisis which uncovered several issues. Among
others, statistical inference is too often viewed as an isolated procedure
limited to the analysis of data that have already been collected. We build on
and further develop an idea proposed by Gelman and Carlin (2014) termed
"prospective and retrospective design analysis". Rather than focusing only on
the statistical significance of a result and on the classical control of type I
and type II errors, a comprehensive design analysis involves reasoning about
what can be considered a plausible effect size. Furthermore, it introduces two
relevant inferential risks: the exaggeration ratio or Type M error (i.e., the
predictable average overestimation of an effect that emerges as statistically
significant), and the sign error or Type S error (i.e., the risk that a
statistically significant effect is estimated in the wrong direction). Another
important aspect of design analysis is that it can be usefully carried out both
in the planning phase of a study and for the evaluation of studies that have
already been conducted, thus increasing researchers' awareness during all
phases of a research project. We use a familiar example in psychology where the
researcher is interested in analyzing the differences between two independent
groups. We examine the case in which the plausible effect size is formalized as
a single value, and propose a method in which uncertainty concerning the
magnitude of the effect is formalized via probability distributions. Through
several examples, we show that even though a design analysis requires big
effort, it has the potential to contribute to planning more robust and
replicable studies. Finally, future developments in the Bayesian framework are
discussed.
| stat |
Subspace Clustering of Very Sparse High-Dimensional Data | In this paper we consider the problem of clustering collections of very short
texts using subspace clustering. This problem arises in many applications such
as product categorisation, fraud detection, and sentiment analysis. The main
challenge lies in the fact that the vectorial representation of short texts is
both high-dimensional, due to the large number of unique terms in the corpus,
and extremely sparse, as each text contains a very small number of words with
no repetition. We propose a new, simple subspace clustering algorithm that
relies on linear algebra to cluster such datasets. Experimental results on
identifying product categories from product names obtained from the US Amazon
website indicate that the algorithm can be competitive against state-of-the-art
clustering algorithms.
| stat |
The Implicit Bias of AdaGrad on Separable Data | We study the implicit bias of AdaGrad on separable linear classification
problems. We show that AdaGrad converges to a direction that can be
characterized as the solution of a quadratic optimization problem with the same
feasible set as the hard SVM problem. We also give a discussion about how
different choices of the hyperparameters of AdaGrad might impact this
direction. This provides a deeper understanding of why adaptive methods do not
seem to have the generalization ability as good as gradient descent does in
practice.
| stat |
Learning finite-dimensional coding schemes with nonlinear reconstruction
maps | This paper generalizes the Maurer--Pontil framework of finite-dimensional
lossy coding schemes to the setting where a high-dimensional random vector is
mapped to an element of a compact set of latent representations in a
lower-dimensional Euclidean space, and the reconstruction map belongs to a
given class of nonlinear maps. Under this setup, which encompasses a broad
class of unsupervised representation learning problems, we establish a
connection to approximate generative modeling under structural constraints
using the tools from the theory of optimal transportation. Next, we consider
problem of learning a coding scheme on the basis of a finite collection of
training samples and present generalization bounds that hold with high
probability. We then illustrate the general theory in the setting where the
reconstruction maps are implemented by deep neural nets.
| stat |
A New Spatial Count Data Model with Time-varying Parameters | Recent crash frequency studies incorporate spatiotemporal correlations, but
these studies have two key limitations: i) none of these studies accounts for
temporal variation in model parameters; and ii) Gibbs sampler suffers from
convergence issues due to non-conjugacy. To address the first limitation, we
propose a new count data model that identifies the underlying temporal patterns
of the regression parameters while simultaneously allowing for time-varying
spatial correlation. The model is also extended to incorporate heterogeneity in
non-temporal parameters across spatial units. We tackle the second shortcoming
by deriving a Gibbs sampler that ensures conditionally conjugate posterior
updates for all model parameters. To this end, we take the advantages of
P\'olya-Gamma data augmentation and forward filtering backward sampling (FFBS)
algorithm. After validating the properties of the Gibbs sampler in a Monte
Carlo study, the advantages of the proposed specification are demonstrated in
an empirical application to uncover relationships between crash frequency
spanning across nine years and pavement characteristics. Model parameters
exhibit practically significant temporal patterns (i.e., temporal instability).
For example, the safety benefits of better pavement ride quality are estimated
to increase over time.
| stat |
Adaptive Configuration Oracle for Online Portfolio Selection Methods | Financial markets are complex environments that produce enormous amounts of
noisy and non-stationary data. One fundamental problem is online portfolio
selection, the goal of which is to exploit this data to sequentially select
portfolios of assets to achieve positive investment outcomes while managing
risks. Various algorithms have been proposed for solving this problem in fields
such as finance, statistics and machine learning, among others. Most of the
methods have parameters that are estimated from backtests for good performance.
Since these algorithms operate on non-stationary data that reflects the
complexity of financial markets, we posit that adaptively tuning these
parameters in an intelligent manner is a remedy for dealing with this
complexity. In this paper, we model the mapping between the parameter space and
the space of performance metrics using a Gaussian process prior. We then
propose an oracle based on adaptive Bayesian optimization for automatically and
adaptively configuring online portfolio selection methods. We test the efficacy
of our solution on algorithms operating on equity and index data from various
markets.
| stat |
Perturbations and Causality in Gaussian Latent Variable Models | Causal inference is a challenging problem with observational data alone. The
task becomes easier when having access to data from perturbing the underlying
system, even when happening in a non-randomized way: this is the setting we
consider, encompassing also latent confounding variables. To identify causal
relations among a collections of covariates and a response variable, existing
procedures rely on at least one of the following assumptions: i) the response
variable remains unperturbed, ii) the latent variables remain unperturbed, and
iii) the latent effects are dense. In this paper, we examine a perturbation
model for interventional data, which can be viewed as a mixed-effects linear
structural causal model, over a collection of Gaussian variables that does not
satisfy any of these conditions. We propose a maximum-likelihood estimator --
dubbed DirectLikelihood -- that exploits system-wide invariances to uniquely
identify the population causal structure from unspecific perturbation data, and
our results carry over to linear structural causal models without requiring
Gaussianity. We illustrate the utility of our framework on synthetic data as
well as real data involving California reservoirs and protein expressions.
| stat |
Blocks as geographic discontinuities: The effect of polling place
assignment on voting | A potential voter must incur a number of costs in order to successfully cast
an in-person ballot, including the costs associated with identifying and
traveling to a polling place. In order to investigate how these costs affect
voting behavior, we introduce two quasi-experimental designs that can be used
to study how the political participation of registered voters is affected by
differences in the relative distance that registrants must travel to their
assigned Election Day polling place and whether their polling place remains at
the same location as in a previous election. Our designs make comparisons of
registrants who live on the same residential block, but are assigned to vote at
different polling places. We find that living farther from a polling place and
being assigned to a new polling place reduce in-person Election Day voting, but
that registrants largely offset for this by casting more early in-person and
mail ballots.
| stat |
The X Factor: A Robust and Powerful Approach to X-chromosome-Inclusive
Whole-genome Association Studies | The X-chromosome is often excluded from genome-wide association studies
because of analytical challenges. Some have been investigated such as the
random, skewed or no X-inactivation model uncertainty. Others have received
little to no attention such as the value in considering non-additive and
gene-sex interaction effects, and the inferential consequence of choosing
different baseline alleles. Here we propose a unified and flexible
regression-based association test for the X-chromosome. We provide theoretical
justifications for its robustness in the presence of various model
uncertainties, as well as for its improved power under certain alternatives
when compared with the existing approaches. For completeness, we also revisit
the autosomes and show that the proposed framework leads to a robust and
sometimes much more powerful test than the standard method. Finally, we provide
supporting evidence by revisiting several published genome-wide association
studies. Supplementary materials for this article are available online.
| stat |
Estimating High-dimensional Covariance and Precision Matrices under
General Missing Dependence | A sample covariance matrix $\boldsymbol{S}$ of completely observed data is
the key statistic in a large variety of multivariate statistical procedures,
such as structured covariance/precision matrix estimation, principal component
analysis, and testing of equality of mean vectors. However, when the data are
partially observed, the sample covariance matrix from the available data is
biased and does not provide valid multivariate procedures. To correct the bias,
a simple adjustment method called inverse probability weighting (IPW) has been
used in previous research, yielding the IPW estimator. The estimator plays the
role of $\boldsymbol{S}$ in the missing data context so that it can be plugged
into off-the-shelf multivariate procedures. However, theoretical properties
(e.g. concentration) of the IPW estimator have been only established under very
simple missing structures; every variable of each sample is independently
subject to missing with equal probability.
We investigate the deviation of the IPW estimator when observations are
partially observed under general missing dependency. We prove the optimal
convergence rate $O_p(\sqrt{\log p / n})$ of the IPW estimator based on the
element-wise maximum norm. We also derive similar deviation results even when
implicit assumptions (known mean and/or missing probability) are relaxed. The
optimal rate is especially crucial in estimating a precision matrix, because of
the "meta-theorem" that claims the rate of the IPW estimator governs that of
the resulting precision matrix estimator. In the simulation study, we discuss
non-positive semi-definiteness of the IPW estimator and compare the estimator
with imputation methods, which are practically important.
| stat |
Complexity of zigzag sampling algorithm for strongly log-concave
distributions | We study the computational complexity of zigzag sampling algorithm for
strongly log-concave distributions. The zigzag process has the advantage of not
requiring time discretization for implementation, and that each proposed
bouncing event requires only one evaluation of partial derivative of the
potential, while its convergence rate is dimension independent. Using these
properties, we prove that the zigzag sampling algorithm achieves $\varepsilon$
error in chi-square divergence with a computational cost equivalent to
$O\bigl(\kappa^2 d^\frac{1}{2}(\log\frac{1}{\varepsilon})^{\frac{3}{2}}\bigr)$
gradient evaluations in the regime $\kappa \ll \frac{d}{\log d}$ under a warm
start assumption, where $\kappa$ is the condition number and $d$ is the
dimension.
| stat |
Theory of Optimal Bayesian Feature Filtering | Optimal Bayesian feature filtering (OBF) is a supervised screening method
designed for biomarker discovery. In this article, we prove two major
theoretical properties of OBF. First, optimal Bayesian feature selection under
a general family of Bayesian models reduces to filtering if and only if the
underlying Bayesian model assumes all features are mutually independent.
Therefore, OBF is optimal if and only if one assumes all features are mutually
independent, and OBF is the only filter method that is optimal under at least
one model in the general Bayesian framework. Second, OBF under independent
Gaussian models is consistent under very mild conditions, including cases where
the data is non-Gaussian with correlated features. This result provides
conditions where OBF is guaranteed to identify the correct feature set given
enough data, and it justifies the use of OBF in non-design settings where its
assumptions are invalid.
| stat |
Functional Optimal Transport: Mapping Estimation and Domain Adaptation
for Functional data | Optimal transport (OT) has generated much recent interest by its capability
of finding mappings that transport mass from one distribution to another, and
found useful roles in machine learning tasks such as unsupervised learning,
domain adaptation and transfer learning. On the other hand, in many
applications data are generated by complex mechanisms involving convoluted
spaces of functions, curves and surfaces in high dimensions. Functional data
analysis provides a useful framework of treatment for such domains. In this
paper we introduce a novel formulation of optimal transport problem in
functional spaces and develop an efficient learning algorithm for finding the
stochastic map between functional domains. We apply our method to synthetic
datasets and study the geometric properties of the transport map. Experiments
on real-world datasets of robot arm trajectories and digit numbers further
demonstrate the effectiveness of our method on applications of domain
adaptation and generative modeling.
| stat |
Calibrating wood products for load duration and rate: A statistical look
at three damage models | Lumber and wood-based products are versatile construction materials that are
susceptible to weakening as a result of applied stresses. To assess the effects
of load duration and rate, experiments have been carried out by applying preset
load profiles to sample specimens. This paper studies these effects via a
damage modeling approach, by considering three models in the literature: the
Gerhards and Foschi accumulated damage models, and a degradation model based on
the gamma process. We present a statistical framework for fitting these models
to failure time data generated by a combination of ramp and constant load
settings, and show how estimation uncertainty can be quantified. The models and
methods are illustrated and compared via a novel analysis of a Hemlock lumber
dataset. Practical usage of the fitted damage models is demonstrated with an
application to long-term reliability prediction under stochastic future
loadings.
| stat |
Group sequential designs for negative binomial outcomes | Count data and recurrent events in clinical trials, such as the number of
lesions in magnetic resonance imaging in multiple sclerosis, the number of
relapses in multiple sclerosis, the number of hospitalizations in heart
failure, and the number of exacerbations in asthma or in chronic obstructive
pulmonary disease (COPD) are often modeled by negative binomial distributions.
In this manuscript we study planning and analyzing clinical trials with group
sequential designs for negative binomial outcomes. We propose a group
sequential testing procedure for negative binomial outcomes based on Wald
statistics using maximum likelihood estimators. The asymptotic distribution of
the proposed group sequential tests statistics are derived. The finite sample
size properties of the proposed group sequential test for negative binomial
outcomes and the methods for planning the respective clinical trials are
assessed in a simulation study. The simulation scenarios are motivated by
clinical trials in chronic heart failure and relapsing multiple sclerosis,
which cover a wide range of practically relevant settings. Our research assures
that the asymptotic normal theory of group sequential designs can be applied to
negative binomial outcomes when the hypotheses are tested using Wald statistics
and maximum likelihood estimators. We also propose two methods, one based on
Student's t-distribution and one based on resampling, to improve type I error
rate control in small samples. The statistical methods studied in this
manuscript are implemented in the R package \textit{gscounts}, which is
available for download on the Comprehensive R Archive Network (CRAN).
| stat |
Calibration of Distributionally Robust Empirical Optimization Models | We study the out-of-sample properties of robust empirical optimization
problems with smooth $\phi$-divergence penalties and smooth concave objective
functions, and develop a theory for data-driven calibration of the non-negative
"robustness parameter" $\delta$ that controls the size of the deviations from
the nominal model. Building on the intuition that robust optimization reduces
the sensitivity of the expected reward to errors in the model by controlling
the spread of the reward distribution, we show that the first-order benefit of
``little bit of robustness" (i.e., $\delta$ small, positive) is a significant
reduction in the variance of the out-of-sample reward while the corresponding
impact on the mean is almost an order of magnitude smaller. One implication is
that substantial variance (sensitivity) reduction is possible at little cost if
the robustness parameter is properly calibrated. To this end, we introduce the
notion of a robust mean-variance frontier to select the robustness parameter
and show that it can be approximated using resampling methods like the
bootstrap. Our examples show that robust solutions resulting from "open loop"
calibration methods (e.g., selecting a $90\%$ confidence level regardless of
the data and objective function) can be very conservative out-of-sample, while
those corresponding to the robustness parameter that optimizes an estimate of
the out-of-sample expected reward (e.g., via the bootstrap) with no regard for
the variance are often insufficiently robust.
| stat |
A multi-surrogate higher-order singular value decomposition tensor
emulator for spatio-temporal simulators | We introduce methodology to construct an emulator for environmental and
ecological spatio-temporal processes that uses the higher order singular value
decomposition (HOSVD) as an extension of singular value decomposition (SVD)
approaches to emulation. Some important advantages of the method are that it
allows for the use of a combination of supervised learning methods (e.g.,
neural networks, random forests, and Gaussian process regression) and also
allows for the prediction of process values at spatial locations and time
points that were not used in the training sample. The method is demonstrated
with two applications: the first is a periodic solution to a shallow ice
approximation partial differential equation from glaciology, and second is an
agent-based model of collective animal movement. In both cases, we demonstrate
the value of combining different machine learning models (i.e., a
multi-surrogate approach) for accurate emulation. In addition, in the
agent-based model case we demonstrate the ability of the tensor emulator to
successfully capture individual behavior in space and time. We demonstrate via
a real data example the ability to perform Bayesian inference in order to learn
parameters governing collective animal behavior.
| stat |
Popper's falsification and corroboration from the statistical
perspectives | The role of probability appears unchallenged as the key measure of
uncertainty, used among other things for practical induction in the empirical
sciences. Yet, Popper was emphatic in his rejection of inductive probability
and of the logical probability of hypotheses; furthermore, for him, the degree
of corroboration cannot be a probability. Instead he proposed a deductive
method of testing. In many ways this dialectic tension has many parallels in
statistics, with the Bayesians on logico-inductive side vs the non-Bayesians or
the frequentists on the other side. Simplistically Popper seems to be on the
frequentist side, but recent synthesis on the non-Bayesian side might direct
the Popperian views to a more nuanced destination. Logical probability seems
perfectly suited to measure partial evidence or support, so what can we use if
we are to reject it? For the past 100 years, statisticians have also developed
a related concept called likelihood, which has played a central role in
statistical modelling and inference. Remarkably, this Fisherian concept of
uncertainty is largely unknown or at least severely under-appreciated in
non-statistical literature. As a measure of corroboration, the likelihood
satisfies the Popperian requirement that it is not a probability. Our aim is to
introduce the likelihood and its recent extension via a discussion of two
well-known logical fallacies in order to highlight that its lack of recognition
may have led to unnecessary confusion in our discourse about falsification and
corroboration of hypotheses. We highlight the 100 years of development of
likelihood concepts. The year 2021 will mark the 100-year anniversary of the
likelihood, so with this paper we wish it a long life and increased
appreciation in non-statistical literature.
| stat |
A fractional Brownian -- Hawkes model for the Italian electricity spot
market: estimation and forecasting | We propose a model for the description and the forecast of the gross prices
of electricity in the liberalized Italian energy market via an additive
two-factor model driven by both a Hawkes and a fractional Brownian processes.
We discuss the seasonality, the identification of spikes and the estimates of
the Hurst coefficient. After the calibration and the validation of the model,
we discuss its forecasting performance via a class of adequate evaluation
metrics.
| stat |
The Power of Graph Convolutional Networks to Distinguish Random Graph
Models: Short Version | Graph convolutional networks (GCNs) are a widely used method for graph
representation learning. We investigate the power of GCNs, as a function of
their number of layers, to distinguish between different random graph models on
the basis of the embeddings of their sample graphs. In particular, the graph
models that we consider arise from graphons, which are the most general
possible parameterizations of infinite exchangeable graph models and which are
the central objects of study in the theory of dense graph limits. We exhibit an
infinite class of graphons that are well-separated in terms of cut distance and
are indistinguishable by a GCN with nonlinear activation functions coming from
a certain broad class if its depth is at least logarithmic in the size of the
sample graph. These results theoretically match empirical observations of
several prior works. Finally, we show a converse result that for pairs of
graphons satisfying a degree profile separation property, a very simple GCN
architecture suffices for distinguishability. To prove our results, we exploit
a connection to random walks on graphs.
| stat |
A principle feature analysis | A key task of data science is to identify relevant features linked to certain
output variables that are supposed to be modeled or predicted. To obtain a
small but meaningful model, it is important to find stochastically independent
variables capturing all the information necessary to model or predict the
output variables sufficiently. Therefore, we introduce in this work a framework
to detect linear and non-linear dependencies between different features. As we
will show, features that are actually functions of other features do not
represent further information. Consequently, a model reduction neglecting such
features conserves the relevant information, reduces noise and thus improves
the quality of the model. Furthermore, a smaller model makes it easier to adopt
a model of a given system. In addition, the approach structures dependencies
within all the considered features. This provides advantages for classical
modeling starting from regression ranging to differential equations and for
machine learning.
To show the generality and applicability of the presented framework 2154
features of a data center are measured and a model for classification for
faulty and non-faulty states of the data center is set up. This number of
features is automatically reduced by the framework to 161 features. The
prediction accuracy for the reduced model even improves compared to the model
trained on the total number of features. A second example is the analysis of a
gene expression data set where from 9513 genes 9 genes are extracted from whose
expression levels two cell clusters of macrophages can be distinguished.
| stat |
Quantifying Observed Prior Impact | We distinguish two questions (i) how much information does the prior contain?
and (ii) what is the effect of the prior? Several measures have been proposed
for quantifying effective prior sample size, for example Clarke [1996] and
Morita et al. [2008]. However, these measures typically ignore the likelihood
for the inference currently at hand, and therefore address (i) rather than
(ii). Since in practice (ii) is of great concern, Reimherr et al. [2014]
introduced a new class of effective prior sample size measures based on
prior-likelihood discordance. We take this idea further towards its natural
Bayesian conclusion by proposing measures of effective prior sample size that
not only incorporate the general mathematical form of the likelihood but also
the specific data at hand. Thus, our measures do not average across datasets
from the working model, but condition on the current observed data.
Consequently, our measures can be highly variable, but we demonstrate that this
is because the impact of a prior can be highly variable. Our measures are Bayes
estimates of meaningful quantities and well communicate the extent to which
inference is determined by the prior, or framed differently, the amount of
effort saved due to having prior information. We illustrate our ideas through a
number of examples including a Gaussian conjugate model (continuous
observations), a Beta-Binomial model (discrete observations), and a linear
regression model (two unknown parameters). Future work on further developments
of the methodology and an application to astronomy are discussed at the end.
| stat |
Efficient nonparametric causal inference with missing exposure
information | Missing exposure information is a very common feature of many observational
studies. Here we study identifiability and efficient estimation of causal
effects on vector outcomes, in such cases where treatment is unconfounded but
partially missing. We consider a missing at random setting where missingness in
treatment can depend not only on complex covariates, but also on post-treatment
outcomes. We give a new identifying expression for average treatment effects in
this setting, along with the efficient influence function for this parameter in
a nonparametric model, which yields a nonparametric efficiency bound. We use
this latter result to construct nonparametric estimators that are less
sensitive to the curse of dimensionality than usual, e.g., by having faster
rates of convergence than the complex nuisance estimators they rely on. Further
we show that these estimators can be root-n consistent and asymptotically
normal under weak nonparametric conditions, even when constructed using
flexible machine learning. Finally we apply these results to the problem of
causal inference with a partially missing instrumental variable.
| stat |
Empirical Bayes mean estimation with nonparametric errors via order
statistic regression | We study empirical Bayes estimation of the effect sizes of $N$ units from $K$
noisy observations on each unit. We show that it is possible to achieve
near-Bayes optimal mean squared error, without any assumptions or knowledge
about the effect size distribution or the noise. The noise distribution can be
heteroskedastic and vary arbitrarily from unit to unit. Our proposal, which we
call Aurora, leverages the replication inherent in the $K$ observations per
unit and recasts the effect size estimation problem as a general regression
problem. Aurora with linear regression provably matches the performance of a
wide array of estimators including the sample mean, the trimmed mean, the
sample median, as well as James-Stein shrunk versions thereof. Aurora automates
effect size estimation for Internet-scale datasets, as we demonstrate on Google
data.
| stat |
The P-T Probability Framework for Semantic Communication, Falsification,
Confirmation, and Bayesian Reasoning | Many researchers want to unify probability and logic by defining logical
probability or probabilistic logic reasonably. This paper tries to unify
statistics and logic so that we can use both statistical probability and
logical probability at the same time. For this purpose, this paper proposes the
P-T probability framework, which is assembled with Shannon's statistical
probability framework for communication, Kolmogorov's probability axioms for
logical probability, and Zadeh's membership functions used as truth functions.
Two kinds of probabilities are connected by an extended Bayes' theorem, with
which we can convert a likelihood function and a truth function from one to
another. Hence, we can train truth functions (in logic) by sampling
distributions (in statistics). This probability framework was developed in the
author's long-term studies on semantic information, statistical learning, and
color vision. This paper first proposes the P-T probability framework and
explains different probabilities in it by its applications to semantic
information theory. Then, this framework and the semantic information methods
are applied to statistical learning, statistical mechanics, hypothesis
evaluation (including falsification), confirmation, and Bayesian reasoning.
Theoretical applications illustrate the reasonability and practicability of
this framework. This framework is helpful for interpretable AI. To interpret
neural networks, we need further study.
| stat |
High-dimensional MANOVA via Bootstrapping and its Application to
Functional and Sparse Count Data | We propose a new approach to the problem of high-dimensional multivariate
ANOVA via bootstrapping max statistics that involve the differences of sample
mean vectors. The proposed method proceeds via the construction of simultaneous
confidence regions for the differences of population mean vectors. It is suited
to simultaneously test the equality of several pairs of mean vectors of
potentially more than two populations. By exploiting the variance decay
property that is a natural feature in relevant applications, we are able to
provide dimension-free and nearly-parametric convergence rates for Gaussian
approximation, bootstrap approximation, and the size of the test. We
demonstrate the proposed approach with ANOVA problems for functional data and
sparse count data. The proposed methodology is shown to work well in
simulations and several real data applications.
| stat |
Analysis of "Learn-As-You-Go" (LAGO) Studies | In learn-as-you-go (LAGO) adaptive studies, the intervention is a complex
package consisting of multiple components, and is adapted in stages during the
study based on past outcome data. This design formalizes standard practice, and
desires for practice, in public health intervention studies. An effective
intervention package is sought, while minimizing intervention package cost.
When analyzing data from a learn-as-you-go study, the interventions in later
stages depend upon the outcomes in the previous stages, violating standard
statistical theory. We develop methods for estimating the intervention effects
in a LAGO study. We prove consistency and asymptotic normality using a novel
coupling argument, ensuring the validity of the test for the hypothesis of no
overall intervention effect. We develop a confidence set for the optimal
intervention package and confidence bands for the success probabilities under
alternative package compositions. We illustrate our methods in the BetterBirth
Study, which aimed to improve maternal and neonatal outcomes among 157,689
births in Uttar Pradesh, India through a complex, multi-component intervention
package.
| stat |
Towards Optimal Estimation of Bivariate Isotonic Matrices with Unknown
Permutations | Many applications, including rank aggregation, crowd-labeling, and graphon
estimation, can be modeled in terms of a bivariate isotonic matrix with unknown
permutations acting on its rows and/or columns. We consider the problem of
estimating an unknown matrix in this class, based on noisy observations of
(possibly, a subset of) its entries. We design and analyze polynomial-time
algorithms that improve upon the state of the art in two distinct metrics,
showing, in particular, that minimax optimal, computationally efficient
estimation is achievable in certain settings. Along the way, we prove matching
upper and lower bounds on the minimax radii of certain cone testing problems,
which may be of independent interest.
| stat |
The use of registry data to extrapolate overall survival results from
randomised controlled trials | Background: Pre-marketing authorisation estimates of survival are generally
restricted to those observed directly in randomised controlled trials (RCTs).
However, for regulatory and Health Technology Assessment (HTA) decision-making
a longer time horizon is often required than is studied in RCTs. Therefore,
extrapolation is required to estimate long-term treatment effect. Registry data
can provide evidence to support extrapolation of treatment effects from RCTs,
which are considered the main sources of evidence of effect for new drug
applications. A number of methods are available to extrapolate survival data,
such as Exponential, Weibull, Gompertz, log-logistic or log-normal parametric
models. The different methods have varying functional forms and can result in
different survival estimates.
Methods: The aim of this paper was to use registry data to supplement the
relatively short term RCT data to obtain long term estimates of effect. No
formal hypotheses were tested. We explore the above parametric regression
models as well as a nonparametric regression model based on local linear
(parametric) regression. We also explore a Bayesian model constrained to the
long term estimate of survival reported in literature, a Bayesian power prior
approach on the variability observed from published literature, and a Bayesian
Model Averaging (BMA) approach. The methods were applied to extrapolate overall
survival of a RCT in metastatic melanoma.
Results: The results showed that the BMA approach was able to fit the RCT
data well, with the lowest variability of the area under the curve up to 72
months with or without the SEER Medicare registry.
Conclusion: the BMA approach is a viable approach to extrapolate overall
survival in the absence of long term data.
| stat |
CRPS Learning | Combination and aggregation techniques can improve forecast accuracy
substantially. This also holds for probabilistic forecasting methods where full
predictive distributions are combined. There are several time-varying and
adaptive weighting schemes like Bayesian model averaging (BMA). However, the
performance of different forecasters may vary not only over time but also in
parts of the distribution. So one may be more accurate in the center of the
distributions, and other ones perform better in predicting the distribution's
tails. Consequently, we introduce a new weighting procedure that considers both
varying performance across time and the distribution. We discuss pointwise
online aggregation algorithms that optimize with respect to the continuous
ranked probability score (CRPS). After analyzing the theoretical properties of
a fully adaptive Bernstein online aggregation (BOA) method, we introduce
smoothing procedures for pointwise CRPS learning. The properties are confirmed
and discussed using simulation studies. Additionally, we illustrate the
performance in a forecasting study for carbon markets. In detail, we predict
the distribution of European emission allowance prices.
| stat |
A Generative Approach to Joint Modeling of Quantitative and Qualitative
Responses | In many scientific areas, data with quantitative and qualitative (QQ)
responses are commonly encountered with a large number of predictors. By
exploring the association between QQ responses, existing approaches often
consider a joint model of QQ responses given the predictor variables. However,
the dependency among predictive variables also provides useful information for
modeling QQ responses. In this work, we propose a generative approach to model
the joint distribution of the QQ responses and predictors. The proposed
generative model provides efficient parameter estimation under a penalized
likelihood framework. It achieves accurate classification for qualitative
response and accurate prediction for quantitative response with efficient
computation. Because of the generative approach framework, the asymptotic
optimality of classification and prediction of the proposed method can be
established under some regularity conditions. The performance of the proposed
method is examined through simulations and real case studies in material
science and genetics.
| stat |
Classification with unknown class-conditional label noise on non-compact
feature spaces | We investigate the problem of classification in the presence of unknown
class-conditional label noise in which the labels observed by the learner have
been corrupted with some unknown class dependent probability. In order to
obtain finite sample rates, previous approaches to classification with unknown
class-conditional label noise have required that the regression function is
close to its extrema on sets of large measure. We shall consider this problem
in the setting of non-compact metric spaces, where the regression function need
not attain its extrema.
In this setting we determine the minimax optimal learning rates (up to
logarithmic factors). The rate displays interesting threshold behaviour: When
the regression function approaches its extrema at a sufficient rate, the
optimal learning rates are of the same order as those obtained in the
label-noise free setting. If the regression function approaches its extrema
more gradually then classification performance necessarily degrades. In
addition, we present an adaptive algorithm which attains these rates without
prior knowledge of either the distributional parameters or the local density.
This identifies for the first time a scenario in which finite sample rates are
achievable in the label noise setting, but they differ from the optimal rates
without label noise.
| stat |
Flexible marked spatio-temporal point processes with applications to
event sequences from association football | We develop a new family of marked point processes by focusing the
characteristic properties of marked Hawkes processes exclusively to the space
of marks, providing the freedom to specify a different model for the occurrence
times. This is possible through a decomposition of the joint distribution of
marks and times that allows to separately specify the conditional distribution
of marks given the filtration of the process and the current time. We develop a
Bayesian framework for the inference and prediction from this family of marked
point processes that can naturally accommodate process and point-specific
covariate information to drive cross-excitations, offering wide flexibility and
applicability in the modelling of real-world processes. The framework is used
here for the modelling of in-game event sequences from association football,
resulting not only in inferences about previously unquantified characteristics
of the game dynamics and extraction of event-specific team abilities, but also
in predictions for the occurrence of events of interest, such as goals, corners
or fouls, in a specified interval of time.
| stat |
Kullback-Leibler-Based Discrete Relative Risk Models for Integration of
Published Prediction Models with New Dataset | Existing literature for prediction of time-to-event data has primarily
focused on risk factors from an individual dataset. However, these analyses may
suffer from small sample sizes, high dimensionality and low signal-to-noise
ratios. To improve prediction stability and better understand risk factors
associated with outcomes of interest, we propose a Kullback-Leibler-based
discrete relative risk modeling procedure. Simulations and real data analysis
are conducted to show the advantage of the proposed methods compared with those
solely based on local dataset or prior models.
| stat |
Individualized Group Learning | Many massive data are assembled through collections of information of a large
number of individuals in a population. The analysis of such data, especially in
the aspect of individualized inferences and solutions, has the potential to
create significant value for practical applications. Traditionally, inference
for an individual in the data set is either solely relying on the information
of the individual or from summarizing the information about the whole
population. However, with the availability of big data, we have the
opportunity, as well as a unique challenge, to make a more effective
individualized inference that takes into consideration of both the population
information and the individual discrepancy. To deal with the possible
heterogeneity within the population while providing effective and credible
inferences for individuals in a data set, this article develops a new approach
called the individualized group learning (iGroup). The iGroup approach uses
local nonparametric techniques to generate an individualized group by pooling
other entities in the population which share similar characteristics with the
target individual, even when individual estimates are biased due to limited
number of observations. Three general cases of iGroup are discussed, and their
asymptotic performances are investigated. Both theoretical results and
empirical simulations reveal that, by applying iGroup, the performance of
statistical inference on the individual level are ensured and can be
substantially improved from inference based on either solely individual
information or entire population information. The method has a broad range of
applications. Two examples in financial statistics and maritime anomaly
detection are presented.
| stat |
Off-policy Learning for Multiple Loggers | It is well known that the historical logs are used for evaluating and
learning policies in interactive systems, e.g. recommendation, search, and
online advertising. Since direct online policy learning usually harms user
experiences, it is more crucial to apply off-policy learning in real-world
applications instead. Though there have been some existing works, most are
focusing on learning with one single historical policy. However, in practice,
usually a number of parallel experiments, e.g. multiple AB tests, are performed
simultaneously. To make full use of such historical data, learning policies
from multiple loggers becomes necessary. Motivated by this, in this paper, we
investigate off-policy learning when the training data coming from multiple
historical policies. Specifically, policies, e.g. neural networks, can be
learned directly from multi-logger data, with counterfactual estimators. In
order to understand the generalization ability of such estimator better, we
conduct generalization error analysis for the empirical risk minimization
problem. We then introduce the generalization error bound as the new risk
function, which can be reduced to a constrained optimization problem. Finally,
we give the corresponding learning algorithm for the new constrained problem,
where we can appeal to the minimax problems to control the constraints.
Extensive experiments on benchmark datasets demonstrate that the proposed
methods achieve better performances than the state-of-the-arts.
| stat |
Strong Sure Screening of Ultra-high Dimensional Data with Interaction
Effects | Ultrahigh dimensional data sets are becoming increasingly prevalent in areas
such as bioinformatics, medical imaging, and social network analysis. Sure
independent screening of such data is commonly used to analyze such data.
Nevertheless, few methods exist for screening for interactions among
predictors. Moreover, extant interaction screening methods prove to be highly
inaccurate when applied to data sets exhibiting strong interactive effects, but
weak marginal effects, on the response. We propose a new interaction screening
procedure based on joint cumulants which is not inhibited by such limitations.
Under a collection of sensible conditions, we demonstrate that our interaction
screening procedure has the strong sure screening property. Four simulations
are used to investigate the performance of our method relative to two other
interaction screening methods. We also apply a two-stage analysis to a real
data example by first employing our proposed method, and then further examining
a subset of selected covariates using multifactor dimensionality reduction.
| stat |
Model Averaging by Cross-validation for Partially Linear Functional
Additive Models | We consider averaging a number of candidate models to produce a prediction of
lower risk in the context of partially linear functional additive models. These
models incorporate the parametric effect of scalar variables and the additive
effect of a functional variable to describe the relationship between response
variable and regressors. We develop a model averaging scheme that assigns the
weights by minimizing a cross-validation criterion. Under the framework of
model misspecification, the resulting estimator is proved to be asymptotically
optimal in terms of the lowest possible square error loss of prediction. Also,
simulation studies and real data analysis demonstrate the good performance of
our proposed method.
| stat |
Theoretical and Practical Limits of Kolmogorov-Zurbenko Periodograms
with DiRienzo-Zurbenko Algorithm Smoothing in the Spectral Analysis of Time
Series Data | The Kolomogorov-Zurbenko periodogram with DiRienzo-Zurbenko algorithm
smoothing is the state-of-the-art method for spectral analysis of time series
data. Because this approach assumes that a sinusoidal model underlies
time-series data and because its algorithms are adaptive in nature, it is
superior to traditional use of autoregressive integral moving average (ARIMA)
algorithms. This article begins with a presentation of its statistical
derivation and development followed by instructions for accessing and utilizing
this approach within the R statistical program platform. The discussion then
turns to a presentation of its theoretical and practical limits with regard to
sensitivity (i.e., ability to detect weak signals), accuracy (i.e., ability to
correctly identify signal frequencies), resolution (i.e., ability to resolve
signals with close frequencies), and robustness with respect to missing data
(i.e., sensitivity and accuracy despite high levels of missingness). Next using
a simulated time series in which two signals close in frequency are embedded in
significant amounts of random noise, the predictive power of this approach is
compared to the traditional ARIMA approach, with support also garnered for its
being robust even in the face of significant levels of missing data. The
article concludes with brief descriptions of studies across a range of
scientific disciplines that have capitalized on the power of the
Kolmogorov-Zurbenko periodogram with DiRienzo-Zurbenko algorithm smoothing.
| stat |
Bias and sensitivity analysis for unmeasured confounders in linear
structural equation models | In this paper, we consider the extent of the biases that may arise when an
unmeasured confounder is omitted from a structural equation model (SEM) and we
propose sensitivity analysis techniques to correct for such biases. We give an
analysis of which effects in an SEM are, and are not, biased by an unmeasured
confounder. It is shown that a single unmeasured confounder will bias not just
one, but numerous, effects in an SEM. We present sensitivity analysis
techniques to correct for biases in total, direct, and indirect effects when
using SEM analyses, and illustrate these techniques with a study of aging and
cognitive function.
| stat |
A connection between the pattern classification problem and the General
Linear Model for statistical inference | A connection between the General Linear Model (GLM) in combination with
classical statistical inference and the machine learning (MLE)-based inference
is described in this paper. Firstly, the estimation of the GLM parameters is
expressed as a Linear Regression Model (LRM) of an indicator matrix, that is,
in terms of the inverse problem of regressing the observations. In other words,
both approaches, i.e. GLM and LRM, apply to different domains, the observation
and the label domains, and are linked by a normalization value at the
least-squares solution. Subsequently, from this relationship we derive a
statistical test based on a more refined predictive algorithm, i.e. the
(non)linear Support Vector Machine (SVM) that maximizes the class margin of
separation, within a permutation analysis. The MLE-based inference employs a
residual score and includes the upper bound to compute a better estimation of
the actual (real) error. Experimental results demonstrate how the parameter
estimations derived from each model resulted in different classification
performances in the equivalent inverse problem. Moreover, using real data the
aforementioned predictive algorithms within permutation tests, including such
model-free estimators, are able to provide a good trade-off between type I
error and statistical power.
| stat |
Stochastic Probabilistic Programs | We introduce the notion of a stochastic probabilistic program and present a
reference implementation of a probabilistic programming facility supporting
specification of stochastic probabilistic programs and inference in them.
Stochastic probabilistic programs allow straightforward specification and
efficient inference in models with nuisance parameters, noise, and
nondeterminism. We give several examples of stochastic probabilistic programs,
and compare the programs with corresponding deterministic probabilistic
programs in terms of model specification and inference. We conclude with
discussion of open research topics and related work.
| stat |
Limitations of Lazy Training of Two-layers Neural Networks | We study the supervised learning problem under either of the following two
models: (1) Feature vectors ${\boldsymbol x}_i$ are $d$-dimensional Gaussians
and responses are $y_i = f_*({\boldsymbol x}_i)$ for $f_*$ an unknown quadratic
function; (2) Feature vectors ${\boldsymbol x}_i$ are distributed as a mixture
of two $d$-dimensional centered Gaussians, and $y_i$'s are the corresponding
class labels. We use two-layers neural networks with quadratic activations, and
compare three different learning regimes: the random features (RF) regime in
which we only train the second-layer weights; the neural tangent (NT) regime in
which we train a linearization of the neural network around its initialization;
the fully trained neural network (NN) regime in which we train all the weights
in the network. We prove that, even for the simple quadratic model of point
(1), there is a potentially unbounded gap between the prediction risk achieved
in these three training regimes, when the number of neurons is smaller than the
ambient dimension. When the number of neurons is larger than the number of
dimensions, the problem is significantly easier and both NT and NN learning
achieve zero risk.
| stat |
Leveraging the Fisher randomization test using confidence distributions:
inference, combination and fusion learning | The flexibility and wide applicability of the Fisher randomization test (FRT)
makes it an attractive tool for assessment of causal effects of interventions
from modern-day randomized experiments that are increasing in size and
complexity. This paper provides a theoretical inferential framework for FRT by
establishing its connection with confidence distributions Such a connection
leads to development of (i) an unambiguous procedure for inversion of FRTs to
generate confidence intervals with guaranteed coverage, (ii) generic and
specific methods to combine FRTs from multiple independent experiments with
theoretical guarantees and (iii) new insights on the effect of size of the
Monte Carlo sample on the results of FRT. Our developments pertain to finite
sample settings but have direct extensions to large samples. Simulations and a
case example demonstrate the benefit of these new developments.
| stat |
Seeing the Wind: Visual Wind Speed Prediction with a Coupled
Convolutional and Recurrent Neural Network | Wind energy resource quantification, air pollution monitoring, and weather
forecasting all rely on rapid, accurate measurement of local wind conditions.
Visual observations of the effects of wind---the swaying of trees and flapping
of flags, for example---encode information regarding local wind conditions that
can potentially be leveraged for visual anemometry that is inexpensive and
ubiquitous. Here, we demonstrate a coupled convolutional neural network and
recurrent neural network architecture that extracts the wind speed encoded in
visually recorded flow-structure interactions of a flag and tree in naturally
occurring wind. Predictions for wind speeds ranging from 0.75-11 m/s showed
agreement with measurements from a cup anemometer on site, with a
root-mean-squared error approaching the natural wind speed variability due to
atmospheric turbulence. Generalizability of the network was demonstrated by
successful prediction of wind speed based on recordings of other flags in the
field and in a controlled wind tunnel test. Furthermore, physics-based scaling
of the flapping dynamics accurately predicts the dependence of the network
performance on the video frame rate and duration.
| stat |
A Bayesian Hidden Semi-Markov Model with Covariate-Dependent State
Duration Parameters for High-Frequency Data from Wearable Devices | Data collected by wearable devices in sports provide valuable information
about an athlete's behavior such as their activity, performance, and ability.
These time series data can be studied with approaches such as hidden Markov and
semi-Markov models (HMM and HSMM) for varied purposes including activity
recognition and event detection. HSMMs extend the HMM by explicitly modeling
the time spent in each state. In a discrete-time HSMM, the duration in each
state can be modeled with a zero-truncated Poisson distribution, where the
duration parameter may be state-specific but constant in time. We extend the
HSMM by allowing the state-specific duration parameters to vary in time and
model them as a function of known covariates derived from the wearable device
and observed over a period of time leading up to a state transition. In
addition, we propose a data subsampling approach given that high-frequency data
from wearable devices can violate the conditional independence assumption of
the HSMM. We apply the model to wearable device data collected on a soccer
referee in a Major League Soccer game. We model the referee's physiological
response to the game demands and identify important time-varying effects of
these demands associated with the duration in each state.
| stat |
Modeling Multivariate Spatial-Temporal Data with Latent Low-Dimensional
Dynamics | High-dimensional multivariate spatial-temporal data arise frequently in a
wide range of applications; however, there are relatively few statistical
methods that can simultaneously deal with spatial, temporal and variable-wise
dependencies in large data sets. In this paper, we propose a new approach to
utilize the correlations in variable, space and time to achieve dimension
reduction and to facilitate spatial/temporal predictions in the
high-dimensional settings. The multivariate spatial-temporal process is
represented as a linear transformation of a lower-dimensional latent factor
process. The spatial dependence structure of the factor process is further
represented non-parametrically in terms of latent empirical orthogonal
functions. The low-dimensional structure is completely unknown in our setting
and is learned entirely from data collected irregularly over space but
regularly over time. We propose innovative estimation and prediction methods
based on the latent low-rank structures. Asymptotic properties of the
estimators and predictors are established. Extensive experiments on synthetic
and real data sets show that, while the dimensions are reduced significantly,
the spatial, temporal and variable-wise covariance structures are largely
preserved. The efficacy of our method is further confirmed by the prediction
performances on both synthetic and real data sets.
| stat |
A robust statistical method for Genome-wide association analysis of
human copy number variation | Conducting genome-wide association studies (GWAS) in copy number variation
(CNV) level is a field where few people involves and little statistical
progresses have been achieved, traditional methods suffer from many problems
such as batch effects, heterogeneity across genome, leading to low power or
high false discovery rate. We develop a new robust method to find
disease-risking regions related to CNV's disproportionately distributed between
case and control samples, even if there are batch effects between them, our
test formula is robust to such effects. We propose a new empirical Bayes rule
to deal with overfitting when estimating parameters during testing, this rule
can be extended to the field of model selection, it can be more efficient
compared with traditional methods when there are too much potential models to
be specified. We also give solid theoretical guarantees for our proposed
method, and demonstrate the effectiveness by simulation and realdata analysis.
| stat |
Nonparametric multimodal regression for circular data | Multimodal regression estimation methods are introduced for regression models
involving circular response and/or covariate. The regression estimators are
based on the maximization of the conditional densities of the response variable
over the covariate. Conditional versions of the mean shift and the circular
mean shift algorithms are used to obtain the regression estimators. The
asymptotic properties of the estimators are studied and the problem of
bandwidth selection is discussed.
| stat |
Statistical Monitoring of the Covariance Matrix in Multivariate
Processes: A Literature Review | Monitoring several correlated quality characteristics of a process is common
in modern manufacturing and service industries. Although a lot of attention has
been paid to monitoring the multivariate process mean, not many control charts
are available for monitoring the covariance matrix. This paper presents a
comprehensive overview of the literature on control charts for monitoring the
covariance matrix in a multivariate statistical process monitoring (MSPM)
framework. It classifies the research that has previously appeared in the
literature. We highlight the challenging areas for research and provide some
directions for future research.
| stat |
Regression by clustering using Metropolis-Hastings | High quality risk adjustment in health insurance markets weakens insurer
incentives to engage in inefficient behavior to attract lower-cost enrollees.
We propose a novel methodology based on Markov Chain Monte Carlo methods to
improve risk adjustment by clustering diagnostic codes into risk groups optimal
for health expenditure prediction. We test the performance of our methodology
against common alternatives using panel data from 500 thousand enrollees of the
Colombian Healthcare System. Results show that our methodology outperforms
common alternatives and suggest that it has potential to improve access to
quality healthcare for the chronically ill.
| stat |
Maximum likelihood estimation of the Fisher-Bingham distribution via
efficient calculation of its normalizing constant | This paper proposes an efficient numerical integration formula to compute the
normalizing constant of Fisher--Bingham distributions. This formula uses a
numerical integration formula with the continuous Euler transform to a
Fourier-type integral representation of the normalizing constant. As this
method is fast and accurate, it can be applied to the calculation of the
normalizing constant of high-dimensional Fisher--Bingham distributions. More
precisely, the error decays exponentially with an increase in the integration
points, and the computation cost increases linearly with the dimensions. In
addition, this formula is useful for calculating the gradient and Hessian
matrix of the normalizing constant. Therefore, we apply this formula to
efficiently calculate the maximum likelihood estimation (MLE) of
high-dimensional data. Finally, we apply the MLE to the hyperspherical
variational auto-encoder (S-VAE), a deep-learning-based generative model that
restricts the latent space to a unit hypersphere. We use the S-VAE trained with
images of handwritten numbers to estimate the distributions of each label. This
application is useful for adding new labels to the models.
| stat |
Strategic Bayesian Asset Allocation | Strategic asset allocation requires an investor to select stocks from a given
basket of assets. The perspective of our investor is to maximize risk-adjusted
alpha returns relative to a benchmark index. Historical returns are used to
provide inputs into an optimization algorithm. Our approach uses Bayesian
regularization to not only provide stock selection but also optimal sequential
portfolio weights. By incorporating investor preferences with a number of
different regularization penalties we extend the approaches of Black (1992) and
Puelz (2015). We tailor standard sparse MCMC algorithms to calculate portfolio
weights and perform selection. We illustrate our methodology on stock selection
from the SP100 stock index and from the top fifty holdings of two hedge funds
Renaissance Technologies and Viking Global. Finally, we conclude with
directions for future research.
| stat |
Convex and Nonconvex Optimization Are Both Minimax-Optimal for Noisy
Blind Deconvolution | We investigate the effectiveness of convex relaxation and nonconvex
optimization in solving bilinear systems of equations (a.k.a. blind
deconvolution under a subspace model). Despite the wide applicability, the
theoretical understanding about these two paradigms remains largely inadequate
in the presence of noise. The current paper makes two contributions by
demonstrating that: (1) convex relaxation achieves minimax-optimal statistical
accuracy vis-\`a-vis random noise, and (2) a two-stage nonconvex algorithm
attains minimax-optimal accuracy within a logarithmic number of iterations.
Both results improve upon the state-of-the-art results by some factors that
scale polynomially in the problem dimension.
| stat |
Global convergence of neuron birth-death dynamics | Neural networks with a large number of parameters admit a mean-field
description, which has recently served as a theoretical explanation for the
favorable training properties of "overparameterized" models. In this regime,
gradient descent obeys a deterministic partial differential equation (PDE) that
converges to a globally optimal solution for networks with a single hidden
layer under appropriate assumptions. In this work, we propose a non-local mass
transport dynamics that leads to a modified PDE with the same minimizer. We
implement this non-local dynamics as a stochastic neuronal birth-death process
and we prove that it accelerates the rate of convergence in the mean-field
limit. We subsequently realize this PDE with two classes of numerical schemes
that converge to the mean-field equation, each of which can easily be
implemented for neural networks with finite numbers of parameters. We
illustrate our algorithms with two models to provide intuition for the
mechanism through which convergence is accelerated.
| stat |
Causality-aware counterfactual confounding adjustment as an alternative
to linear residualization in anticausal prediction tasks based on linear
learners | Linear residualization is a common practice for confounding adjustment in
machine learning (ML) applications. Recently, causality-aware predictive
modeling has been proposed as an alternative causality-inspired approach for
adjusting for confounders. The basic idea is to simulate counterfactual data
that is free from the spurious associations generated by the observed
confounders. In this paper, we compare the linear residualization approach
against the causality-aware confounding adjustment in anticausal prediction
tasks, and show that the causality-aware approach tends to (asymptotically)
outperform the residualization adjustment in terms of predictive performance in
linear learners. Importantly, our results still holds even when the true model
is not linear. We illustrate our results in both regression and classification
tasks, where we compared the causality-aware and residualization approaches
using mean squared errors and classification accuracy in synthetic data
experiments where the linear regression model is mispecified, as well as, when
the linear model is correctly specified. Furthermore, we illustrate how the
causality-aware approach is more stable than residualization with respect to
dataset shifts in the joint distribution of the confounders and outcome
variables.
| stat |
Graphical Lasso and Thresholding: Equivalence and Closed-form Solutions | Graphical Lasso (GL) is a popular method for learning the structure of an
undirected graphical model, which is based on an $l_1$ regularization
technique. The objective of this paper is to compare the computationally-heavy
GL technique with a numerically-cheap heuristic method that is based on simply
thresholding the sample covariance matrix. To this end, two notions of
sign-consistent and inverse-consistent matrices are developed, and then it is
shown that the thresholding and GL methods are equivalent if: (i) the
thresholded sample covariance matrix is both sign-consistent and
inverse-consistent, and (ii) the gap between the largest thresholded and the
smallest un-thresholded entries of the sample covariance matrix is not too
small. By building upon this result, it is proved that the GL method---as a
conic optimization problem---has an explicit closed-form solution if the
thresholded sample covariance matrix has an acyclic structure. This result is
then generalized to arbitrary sparse support graphs, where a formula is found
to obtain an approximate solution of GL. Furthermore, it is shown that the
approximation error of the derived explicit formula decreases exponentially
fast with respect to the length of the minimum-length cycle of the sparsity
graph. The developed results are demonstrated on synthetic data, functional MRI
data, traffic flows for transportation networks, and massive randomly generated
data sets. We show that the proposed method can obtain an accurate
approximation of the GL for instances with the sizes as large as $80,000\times
80,000$ (more than 3.2 billion variables) in less than 30 minutes on a standard
laptop computer running MATLAB, while other state-of-the-art methods do not
converge within 4 hours.
| stat |
Time-based analysis of the NBA hot hand fallacy | The debate surrounding the hot hand in the NBA has been ongoing for many
years. However, many of the previous works on this theme has focused on only
the very next sequential shot attempt, often on very select players. This work
looks in more detail the effect of a made or missed shot on the next series of
shots over a two-year span, with time between shots shown to be a critical
factor in the analysis. Also, multi-year streakiness is analyzed, and all
indications are that players cannot really sustain their good (or bad) fortune
from year to year.
| stat |
Bidirectional Inference Networks: A Class of Deep Bayesian Networks for
Health Profiling | We consider the problem of inferring the values of an arbitrary set of
variables (e.g., risk of diseases) given other observed variables (e.g.,
symptoms and diagnosed diseases) and high-dimensional signals (e.g., MRI images
or EEG). This is a common problem in healthcare since variables of interest
often differ for different patients. Existing methods including Bayesian
networks and structured prediction either do not incorporate high-dimensional
signals or fail to model conditional dependencies among variables. To address
these issues, we propose bidirectional inference networks (BIN), which stich
together multiple probabilistic neural networks, each modeling a conditional
dependency. Predictions are then made via iteratively updating variables using
backpropagation (BP) to maximize corresponding posterior probability.
Furthermore, we extend BIN to composite BIN (CBIN), which involves the
iterative prediction process in the training stage and improves both accuracy
and computational efficiency by adaptively smoothing the optimization
landscape. Experiments on synthetic and real-world datasets (a sleep study and
a dermatology dataset) show that CBIN is a single model that can achieve
state-of-the-art performance and obtain better accuracy in most inference tasks
than multiple models each specifically trained for a different task.
| stat |
Matrix Completion under Low-Rank Missing Mechanism | Matrix completion is a modern missing data problem where both the missing
structure and the underlying parameter are high dimensional. Although missing
structure is a key component to any missing data problems, existing matrix
completion methods often assume a simple uniform missing mechanism. In this
work, we study matrix completion from corrupted data under a novel low-rank
missing mechanism. The probability matrix of observation is estimated via a
high dimensional low-rank matrix estimation procedure, and further used to
complete the target matrix via inverse probabilities weighting. Due to both
high dimensional and extreme (i.e., very small) nature of the true probability
matrix, the effect of inverse probability weighting requires careful study. We
derive optimal asymptotic convergence rates of the proposed estimators for both
the observation probabilities and the target matrix.
| stat |
A Kalman particle filter for online parameter estimation with
applications to affine models | In this paper we address the problem of estimating the posterior distribution
of the static parameters of a continuous time state space model with discrete
time observations by an algorithm that combines the Kalman filter and a
particle filter. The proposed algorithm is semi-recursive and has a two layer
structure, in which the outer layer provides the estimation of the posterior
distribution of the unknown parameters and the inner layer provides the
estimation of the posterior distribution of the state variables. This algorithm
has a similar structure as the so-called recursive nested particle filter, but
unlike the latter filter, in which both layers use a particle filter, this
proposed algorithm introduces a dynamic kernel to sample the parameter
particles in the outer layer to obtain a higher convergence speed. Moreover,
this algorithm also implements the Kalman filter in the inner layer to reduce
the computational time. This algorithm can also be used to estimate the
parameters that suddenly change value. We prove that, for a state space model
with a certain structure, the estimated posterior distribution of the unknown
parameters and the state variables converge to the actual distribution in $L_p$
with rate of order $\mathcal{O}(N^{-\frac{1}{2}}+\delta^{\frac{1}{2}})$, where
$N$ is the number of particles for the parameters in the outer layer and
$\delta$ is the maximum time step between two consecutive observations. We
present numerical results of the implementation of this algorithm, in
particularly we implement this algorithm for affine interest models, possibly
with stochastic volatility, although the algorithm can be applied to a much
broader class of models.
| stat |
Towards Robust and Stable Deep Learning Algorithms for Forward Backward
Stochastic Differential Equations | Applications in quantitative finance such as optimal trade execution, risk
management of options, and optimal asset allocation involve the solution of
high dimensional and nonlinear Partial Differential Equations (PDEs). The
connection between PDEs and systems of Forward-Backward Stochastic Differential
Equations (FBSDEs) enables the use of advanced simulation techniques to be
applied even in the high dimensional setting. Unfortunately, when the
underlying application contains nonlinear terms, then classical methods both
for simulation and numerical methods for PDEs suffer from the curse of
dimensionality. Inspired by the success of deep learning, several researchers
have recently proposed to address the solution of FBSDEs using deep learning.
We discuss the dynamical systems point of view of deep learning and compare
several architectures in terms of stability, generalization, and robustness. In
order to speed up the computations, we propose to use a multilevel
discretization technique. Our preliminary results suggest that the multilevel
discretization method improves solutions times by an order of magnitude
compared to existing methods without sacrificing stability or robustness.
| stat |
Permutation tests under a rotating sampling plan with clustered data | Consider a population consisting of clusters of sampling units, evolving
temporally, spatially, or according to other dynamics. We wish to monitor the
evolution of its means, medians, or other parameters. For administrative
convenience and informativeness, clustered data are often collected via a
rotating plan. Under rotating plans, the observations in the same clusters are
correlated, and observations on the same unit collected on different occasions
are also correlated. Ignoring this correlation structure may lead to invalid
inference procedures. Accommodating cluster structure in parametric models is
difficult or will have a high level of misspecification risk. In this paper, we
explore exchangeability in clustered data collected via a rotating sampling
plan to develop a permutation scheme for testing various hypotheses of
interest. We also introduce a semiparametric density ratio model to facilitate
the multiple population structure in rotating sampling plans. The combination
ensures the validity of the inference methods while extracting maximum
information from the sampling plan. A simulation study indicates that the
proposed tests firmly control the type I error whether or not the data are
clustered. The use of the density ratio model improves the power of the tests.
| stat |
What do adoption patterns of solar panels observed so far tell about
governments' incentive? insight from diffusion models | The paper uses diffusion models to understand the main determinants of
diffusion of solar photovoltaic panels (SPP) worldwide, focusing on the role of
public incentives. We applied the generalized Bass model (GBM) to adoption data
of 26 countries between 1992-2016. The SPP market appears as a frail and
complicate one, lacking public media support. Even the major shocks in adoption
curves, following state incentive implemented after 2006, failed to go beyond
short-term effects and therefore were unable to provide sustained momentum to
the market. This suggests that further barriers to adoption should be removed.
| stat |
Sparse Network Estimation for Dynamical Spatio-temporal Array Models | Neural field models represent neuronal communication on a population level
via synaptic weight functions. Using voltage sensitive dye (VSD) imaging it is
possible to obtain measurements of neural fields with a relatively high spatial
and temporal resolution. The synaptic weight functions represent functional
connectivity in the brain and give rise to a spatio-temporal dependence
structure. We present a stochastic functional differential equation for
modeling neural fields, which leads to a vector autoregressive model of the
data via basis expansions of the synaptic weight functions and time and space
discretization. Fitting the model to data is a pratical challenge as this
represents a large scale regression problem. By using a 1-norm penalty in
combination with localized basis functions it is possible to learn a sparse
network representation of the functional connectivity of the brain, but still,
the explicit construction of a design matrix can be computationally
prohibitive. We demonstrate that by using tensor product basis expansions, the
computation of the penalized estimator via a proximal gradient algorithm
becomes feasible. It is crucial for the computations that the data is organized
in an array as is the case for the three dimensional VSD imaging data. This
allows for the use of array arithmetic that is both memory and time
efficient.The proposed method is implemented and showcased in the R package
dynamo available from CRAN.
| stat |
Individually Fair Ranking | We develop an algorithm to train individually fair learning-to-rank (LTR)
models. The proposed approach ensures items from minority groups appear
alongside similar items from majority groups. This notion of fair ranking is
based on the definition of individual fairness from supervised learning and is
more nuanced than prior fair LTR approaches that simply ensure the ranking
model provides underrepresented items with a basic level of exposure. The crux
of our method is an optimal transport-based regularizer that enforces
individual fairness and an efficient algorithm for optimizing the regularizer.
We show that our approach leads to certifiably individually fair LTR models and
demonstrate the efficacy of our method on ranking tasks subject to demographic
biases.
| stat |
Large scale analysis of generalization error in learning using margin
based classification methods | Large-margin classifiers are popular methods for classification. We derive
the asymptotic expression for the generalization error of a family of
large-margin classifiers in the limit of both sample size $n$ and dimension $p$
going to $\infty$ with fixed ratio $\alpha=n/p$. This family covers a broad
range of commonly used classifiers including support vector machine, distance
weighted discrimination, and penalized logistic regression. Our result can be
used to establish the phase transition boundary for the separability of two
classes. We assume that the data are generated from a single multivariate
Gaussian distribution with arbitrary covariance structure. We explore two
special choices for the covariance matrix: spiked population model and two
layer neural networks with random first layer weights. The method we used for
deriving the closed-form expression is from statistical physics known as the
replica method. Our asymptotic results match simulations already when $n,p$ are
of the order of a few hundreds. For two layer neural networks, we reproduce the
recently developed `double descent' phenomenology for several classification
models. We also discuss some statistical insights that can be drawn from these
analysis.
| stat |
Learning distant cause and effect using only local and immediate credit
assignment | We present a recurrent neural network memory that uses sparse coding to
create a combinatoric encoding of sequential inputs. Using several examples, we
show that the network can associate distant causes and effects in a discrete
stochastic process, predict partially-observable higher-order sequences, and
enable a DQN agent to navigate a maze by giving it memory. The network uses
only biologically-plausible, local and immediate credit assignment. Memory
requirements are typically one order of magnitude less than existing LSTM, GRU
and autoregressive feed-forward sequence learning models. The most significant
limitation of the memory is generalization to unseen input sequences. We
explore this limitation by measuring next-word prediction perplexity on the
Penn Treebank dataset.
| stat |
A Contextual Bandit Bake-off | Contextual bandit algorithms are essential for solving many real-world
interactive machine learning problems. Despite multiple recent successes on
statistically and computationally efficient methods, the practical behavior of
these algorithms is still poorly understood. We leverage the availability of
large numbers of supervised learning datasets to empirically evaluate
contextual bandit algorithms, focusing on practical methods that learn by
relying on optimization oracles from supervised learning. We find that a recent
method (Foster et al., 2018) using optimism under uncertainty works the best
overall. A surprisingly close second is a simple greedy baseline that only
explores implicitly through the diversity of contexts, followed by a variant of
Online Cover (Agarwal et al., 2014) which tends to be more conservative but
robust to problem specification by design. Along the way, we also evaluate
various components of contextual bandit algorithm design such as loss
estimators. Overall, this is a thorough study and review of contextual bandit
methodology.
| stat |
A reckless guide to P-values: local evidence, global errors | This chapter demystifies P-values, hypothesis tests and significance tests,
and introduces the concepts of local evidence and global error rates. The local
evidence is embodied in \textit{this} data and concerns the hypotheses of
interest for \textit{this} experiment, whereas the global error rate is a
property of the statistical analysis and sampling procedure. It is shown using
simple examples that local evidence and global error rates can be, and should
be, considered together when making inferences. Power analysis for experimental
design for hypothesis testing are explained, along with the more locally
focussed expected P-values. Issues relating to multiple testing, HARKing, and
P-hacking are explained, and it is shown that, in many situation, their effects
on local evidence and global error rates are in conflict, a conflict that can
always be overcome by a fresh dataset from replication of key experiments.
Statistics is complicated, and so is science. There is no singular right way to
do either, and universally acceptable compromises may not exist. Statistics
offers a wide array of tools for assisting with scientific inference by
calibrating uncertainty, but statistical inference is not a substitute for
scientific inference. P-values are useful indices of evidence and deserve their
place in the statistical toolbox of basic pharmacologists.
| stat |
Can we trust the standardized mortality ratio? A formal analysis and
evaluation based on axiomatic requirements | Background: The standardized mortality ratio (SMR) is often used to assess
and compare hospital performance. While it has been recognized that hospitals
may differ in their SMRs due to differences in patient composition, there is a
lack of rigorous analysis of this and other - largely unrecognized - properties
of the SMR. Methods: This paper proposes five axiomatic requirements for
adequate standardized mortality measures: strict monotonicity, case-mix
insensitivity, scale insensitivity, equivalence principle, and dominance
principle. Given these axiomatic requirements, effects of variations in patient
composition, hospital size, and actual and expected mortality rates on the SMR
were examined using basic algebra and calculus. In this regard, we
distinguished between standardization using expected mortality rates derived
from a different dataset (external standardization) and standardization based
on a dataset including the considered hospitals (internal standardization).
Results: Under external standardization, the SMR fulfills the axiomatic
requirements of strict monotonicity and scale insensitivity but violates the
requirement of case-mix insensitivity, the equivalence principle, and the
dominance principle. All axiomatic requirements not fulfilled under external
standardization are also not fulfilled under internal standardization. In
addition, the SMR under internal standardization is scale sensitive and
violates the axiomatic requirement of strict monotonicity. Conclusions: The SMR
fulfills only two (none) out of the five proposed axiomatic requirements under
external (internal) standardization. Generally, the SMRs of hospitals are
differently affected by variations in case mix and actual and expected
mortality rates unless the hospitals are identical in these characteristics.
These properties hamper valid assessment and comparison of hospital performance
based on the SMR.
| stat |
Online Batch Decision-Making with High-Dimensional Covariates | We propose and investigate a class of new algorithms for sequential decision
making that interacts with \textit{a batch of users} simultaneously instead of
\textit{a user} at each decision epoch. This type of batch models is motivated
by interactive marketing and clinical trial, where a group of people are
treated simultaneously and the outcomes of the whole group are collected before
the next stage of decision. In such a scenario, our goal is to allocate a batch
of treatments to maximize treatment efficacy based on observed high-dimensional
user covariates. We deliver a solution, named \textit{Teamwork LASSO Bandit
algorithm}, that resolves a batch version of explore-exploit dilemma via
switching between teamwork stage and selfish stage during the whole decision
process. This is made possible based on statistical properties of LASSO
estimate of treatment efficacy that adapts to a sequence of batch observations.
In general, a rate of optimal allocation condition is proposed to delineate the
exploration and exploitation trade-off on the data collection scheme, which is
sufficient for LASSO to identify the optimal treatment for observed user
covariates. An upper bound on expected cumulative regret of the proposed
algorithm is provided.
| stat |
Estimating the Prediction Performance of Spatial Models via Spatial
k-Fold Cross Validation | In machine learning one often assumes the data are independent when
evaluating model performance. However, this rarely holds in practise.
Geographic information data sets are an example where the data points have
stronger dependencies among each other the closer they are geographically. This
phenomenon known as spatial autocorrelation (SAC) causes the standard cross
validation (CV) methods to produce optimistically biased prediction performance
estimates for spatial models, which can result in increased costs and accidents
in practical applications. To overcome this problem we propose a modified
version of the CV method called spatial k-fold cross validation (SKCV), which
provides a useful estimate for model prediction performance without optimistic
bias due to SAC. We test SKCV with three real world cases involving open
natural data showing that the estimates produced by the ordinary CV are up to
40% more optimistic than those of SKCV. Both regression and classification
cases are considered in our experiments. In addition, we will show how the SKCV
method can be applied as a criterion for selecting data sampling density for
new research area.
| stat |
Joint Estimation of Location and Scatter in Complex Elliptical
Distributions: A robust semiparametric and computationally efficient
$R$-estimator of the shape matrix | The joint estimation of the location vector and the shape matrix of a set of
independent and identically Complex Elliptically Symmetric (CES) distributed
observations is investigated from both the theoretical and computational
viewpoints. This joint estimation problem is framed in the original context of
semiparametric models allowing us to handle the (generally unknown) density
generator as an \textit{infinite-dimensional} nuisance parameter. In the first
part of the paper, a computationally efficient and memory saving implementation
of the robust and semiparmaetric efficient $R$-estimator for shape matrices is
derived. Building upon this result, in the second part, a joint estimator,
relying on the Tyler's $M$-estimator of location and on the $R$-estimator of
shape matrix, is proposed and its Mean Squared Error (MSE) performance compared
with the Semiparametric Cram\'{e}r-Rao Bound (CSCRB).
| stat |
Bayesian Pool-based Active Learning With Abstention Feedbacks | We study pool-based active learning with abstention feedbacks, where a
labeler can abstain from labeling a queried example with some unknown
abstention rate. This is an important problem with many useful applications. We
take a Bayesian approach to the problem and develop two new greedy algorithms
that learn both the classification problem and the unknown abstention rate at
the same time. These are achieved by simply incorporating the estimated
abstention rate into the greedy criteria. We prove that both of our algorithms
have near-optimality guarantees: they respectively achieve a
${(1-\frac{1}{e})}$ constant factor approximation of the optimal expected or
worst-case value of a useful utility function. Our experiments show the
algorithms perform well in various practical scenarios.
| stat |
Using Contextual Information to Improve Blood Glucose Prediction | Blood glucose value prediction is an important task in diabetes management.
While it is reported that glucose concentration is sensitive to social context
such as mood, physical activity, stress, diet, alongside the influence of
diabetes pathologies, we need more research on data and methodologies to
incorporate and evaluate signals about such temporal context into prediction
models. Person-generated data sources, such as actively contributed surveys as
well as passively mined data from social media offer opportunity to capture
such context, however the self-reported nature and sparsity of such data mean
that such data are noisier and less specific than physiological measures such
as blood glucose values themselves. Therefore, here we propose a Gaussian
Process model to both address these data challenges and combine blood glucose
and latent feature representations of contextual data for a novel multi-signal
blood glucose prediction task. We find this approach outperforms common methods
for multi-variate data, as well as using the blood glucose values in isolation.
Given a robust evaluation across two blood glucose datasets with different
forms of contextual information, we conclude that multi-signal Gaussian
Processes can improve blood glucose prediction by using contextual information
and may provide a significant shift in blood glucose prediction research and
practice.
| stat |
Getting Better from Worse: Augmented Bagging and a Cautionary Tale of
Variable Importance | As the size, complexity, and availability of data continues to grow,
scientists are increasingly relying upon black-box learning algorithms that can
often provide accurate predictions with minimal a priori model specifications.
Tools like random forests have an established track record of off-the-shelf
success and even offer various strategies for analyzing the underlying
relationships among variables. Here, motivated by recent insights into random
forest behavior, we introduce the simple idea of augmented bagging (AugBagg), a
procedure that operates in an identical fashion to classical bagging and random
forests, but which operates on a larger, augmented space containing additional
randomly generated noise features. Surprisingly, we demonstrate that this
simple act of including extra noise variables in the model can lead to dramatic
improvements in out-of-sample predictive accuracy, sometimes outperforming even
an optimally tuned traditional random forest. As a result, intuitive notions of
variable importance based on improved model accuracy may be deeply flawed, as
even purely random noise can routinely register as statistically significant.
Numerous demonstrations on both real and synthetic data are provided along with
a proposed solution.
| stat |
Inhomogeneous Markov Survival Regression Models | We propose new regression models in survival analysis based on homogeneous
and inhomogeneous phase-type distributions. The intensity function in this
setting plays the role of the hazard function. For unidimensional intensity
matrices, we recover the proportional hazard and accelerated failure time
models, among others. However, when considering higher dimensions, the proposed
methods are only asymptotically equivalent to their classical counterparts and
enjoy greater flexibility in the body of the distribution. For their
estimation, the latent path representation of semi-Markov models is exploited.
Consequently, an adapted EM algorithm is provided and the likelihood is shown
to increase at each iteration. We provide several examples of practical
significance and outline relevant extensions. The practical feasibility of the
models is illustrated on simulated and real-world datasets.
| stat |
Identification of taxon through fuzzy classification | Identification of taxa can be significantly assisted by statistical
classification in two major ways. With a collection of subjects with common
traits measured, it is possible to determine combinations of trait measurements
that signify each taxon in question. These decision regions also make it
possible to classify new observations. In this paper we present a general
Bayesian approach for classifying observations based on traits, whose
measurements follow some (latent) multivariate Gaussian distribution, by
analysis of an original, example data set. Decision rules based on supervised
learning and blockwise Gibbs sampling are presented that either predict a
specific category or fuzzy versions that rather predict a set of categories
containing the most probable categories. This fuzzy discriminant analysis
employs a unified framework of set valued reward functions. We also present a
way of safeguarding for outlying new observations, based on a multivariate type
of $p$-values. Finally, we incorporate model selection through cross-validation
on another original data set.
| stat |
Statistical Molecule Counting in Super-Resolution Fluorescence
Microscopy: Towards Quantitative Nanoscopy | Super-resolution microscopy is rapidly gaining importance as an analytical
tool in the life sciences. A compelling feature is the ability to label
biological units of interest with fluorescent markers in living cells and to
observe them with considerably higher resolution than conventional microscopy
permits. The images obtained this way, however, lack an absolute intensity
scale in terms of numbers of fluorophores observed. We provide an elaborate
model to estimate this information from the raw data. To this end we model the
entire process of photon generation in the fluorophore, their passage trough
the microscope, detection and photo electron amplification in the camera, and
extraction of time series from the microscopic images. At the heart of these
modeling steps is a careful description of the fluorophore dynamics by a novel
hidden Markov model that operates on two time scales (HTMM). Besides the
fluorophore number, information about the kinetic transition rates of the
fluorophore's internal states is also inferred during estimation. We comment on
computational issues that arise when applying our model to simulated or
measured fluorescence traces and illustrate our methodology on simulated data.
| stat |
Robust Model Selection for Finite Mixture of Regression Models Through
Trimming | In this article, we introduce a new variable selection technique through
trimming for finite mixture of regression models. Compared to the traditional
variable selection techniques, the new method is robust and not sensitive to
outliers. The estimation algorithm is introduced and numerical studies are
conducted to examine the finite sample performance of the proposed procedure
and to compare it with other existing methods.
| stat |
Causal Mediation Analysis for Longitudinal Mediators and Survival
Outcomes | Causal mediation analysis studies how the treatment effect of an exposure on
outcomes is mediated through intermediate variables. Although many applications
involve longitudinal data, the existing methods are not directly applicable to
settings where the mediators are measured on irregular time grids. In this
paper, we propose a causal mediation method that accommodates longitudinal
mediators on arbitrary time grids and survival outcomes simultaneously. We take
a functional data analysis perspective and view longitudinal mediators as
realizations of underlying smooth stochastic processes. We define causal
estimands of direct and indirect effects accordingly and provide corresponding
identification assumptions. We employ a functional principal component analysis
approach to estimate the mediator process, and propose a Cox hazard model for
the survival outcome that flexibly adjusts the mediator process. We then derive
a g-computation formula to express the causal estimands using the model
coefficients. The proposed method is applied to a longitudinal data set from
the Amboseli Baboon Research Project to investigate the causal relationships
between early adversity, adult physiological stress responses, and survival
among wild female baboons. We find that adversity experienced in early life has
a significant direct effect on females' life expectancy and survival
probability, but find little evidence that these effects were mediated by
markers of the stress response in adulthood. We further developed a sensitivity
analysis method to assess the impact of potential violation to the key
assumption of sequential ignorability.
| stat |
Bayesian classification, anomaly detection, and survival analysis using
network inputs with application to the microbiome | While the study of a single network is well-established, technological
advances now allow for the collection of multiple networks with relative ease.
Increasingly, anywhere from several to thousands of networks can be created
from brain imaging, gene co-expression data, or microbiome measurements. And
these networks, in turn, are being looked to as potentially powerful features
to be used in modeling. However, with networks being non-Euclidean in nature,
how best to incorporate them into standard modeling tasks is not obvious. In
this paper, we propose a Bayesian modeling framework that provides a unified
approach to binary classification, anomaly detection, and survival analysis
with network inputs. We encode the networks in the kernel of a Gaussian process
prior via their pairwise differences and we discuss several choices of provably
positive definite kernel that can be plugged into our models. Although our
methods are widely applicable, we are motivated here in particular by
microbiome research (where network analysis is emerging as the standard
approach for capturing the interconnectedness of microbial taxa across both
time and space) and its potential for reducing preterm delivery and improving
personalization of prenatal care.
| stat |
Dynamic time series clustering via volatility change-points | This note outlines a method for clustering time series based on a statistical
model in which volatility shifts at unobserved change-points. The model
accommodates some classical stylized features of returns and its relation to
GARCH is discussed. Clustering is performed using a probability metric
evaluated between posterior distributions of the most recent change-point
associated with each series. This implies series are grouped together at a
given time if there is evidence the most recent shifts in their respective
volatilities were coincident or closely timed. The clustering method is
dynamic, in that groupings may be updated in an online manner as data arrive.
Numerical results are given analyzing daily returns of constituents of the S&P
500.
| stat |
Pitfalls of In-Domain Uncertainty Estimation and Ensembling in Deep
Learning | Uncertainty estimation and ensembling methods go hand-in-hand. Uncertainty
estimation is one of the main benchmarks for assessment of ensembling
performance. At the same time, deep learning ensembles have provided
state-of-the-art results in uncertainty estimation. In this work, we focus on
in-domain uncertainty for image classification. We explore the standards for
its quantification and point out pitfalls of existing metrics. Avoiding these
pitfalls, we perform a broad study of different ensembling techniques. To
provide more insight in this study, we introduce the deep ensemble equivalent
score (DEE) and show that many sophisticated ensembling techniques are
equivalent to an ensemble of only few independently trained networks in terms
of test performance.
| stat |
Evaluating Overfit and Underfit in Models of Network Community Structure | A common data mining task on networks is community detection, which seeks an
unsupervised decomposition of a network into structural groups based on
statistical regularities in the network's connectivity. Although many methods
exist, the No Free Lunch theorem for community detection implies that each
makes some kind of tradeoff, and no algorithm can be optimal on all inputs.
Thus, different algorithms will over or underfit on different inputs, finding
more, fewer, or just different communities than is optimal, and evaluation
methods that use a metadata partition as a ground truth will produce misleading
conclusions about general accuracy. Here, we present a broad evaluation of over
and underfitting in community detection, comparing the behavior of 16
state-of-the-art community detection algorithms on a novel and structurally
diverse corpus of 406 real-world networks. We find that (i) algorithms vary
widely both in the number of communities they find and in their corresponding
composition, given the same input, (ii) algorithms can be clustered into
distinct high-level groups based on similarities of their outputs on real-world
networks, and (iii) these differences induce wide variation in accuracy on link
prediction and link description tasks. We introduce a new diagnostic for
evaluating overfitting and underfitting in practice, and use it to roughly
divide community detection methods into general and specialized learning
algorithms. Across methods and inputs, Bayesian techniques based on the
stochastic block model and a minimum description length approach to
regularization represent the best general learning approach, but can be
outperformed under specific circumstances. These results introduce both a
theoretically principled approach to evaluate over and underfitting in models
of network community structure and a realistic benchmark by which new methods
may be evaluated and compared.
| stat |
Scalable Nonparametric Sampling from Multimodal Posteriors with the
Posterior Bootstrap | Increasingly complex datasets pose a number of challenges for Bayesian
inference. Conventional posterior sampling based on Markov chain Monte Carlo
can be too computationally intensive, is serial in nature and mixes poorly
between posterior modes. Further, all models are misspecified, which brings
into question the validity of the conventional Bayesian update. We present a
scalable Bayesian nonparametric learning routine that enables posterior
sampling through the optimization of suitably randomized objective functions. A
Dirichlet process prior on the unknown data distribution accounts for model
misspecification, and admits an embarrassingly parallel posterior bootstrap
algorithm that generates independent and exact samples from the nonparametric
posterior distribution. Our method is particularly adept at sampling from
multimodal posterior distributions via a random restart mechanism. We
demonstrate our method on Gaussian mixture model and sparse logistic regression
examples.
| stat |
A Multi-Site Stochastic Weather Generator for High-Frequency
Precipitation Using Censored Skew-Symmetric Distribution | Stochastic weather generators (SWGs) are digital twins of complex weather
processes and widely used in agriculture and urban design. Due to improved
measuring instruments, an accurate SWG for high-frequency precipitation is now
possible. However, high-frequency precipitation data are more zero-inflated,
skewed, and heavy-tailed than common (hourly or daily) precipitation data.
Therefore, classical methods that either model precipitation occurrence
independently of their intensity or assume that the precipitation follows a
censored meta-Gaussian process may not be appropriate. In this work, we propose
a novel multi-site precipitation generator that drives both occurrence and
intensity by a censored non-Gaussian vector autoregression model with
skew-symmetric dynamics. The proposed SWG is advantageous in modeling skewed
and heavy-tailed data with direct physical and statistical interpretations. We
apply the proposed model to 30-second precipitation based on the data obtained
from a dense gauge network in Lausanne, Switzerland. In addition to reproducing
the high-frequency precipitation, the model can provide accurate predictions as
the long short-term memory (LSTM) network but with uncertainties and more
interpretable results.
| stat |
Sharpened Generalization Bounds based on Conditional Mutual Information
and an Application to Noisy, Iterative Algorithms | The information-theoretic framework of Russo and J. Zou (2016) and Xu and
Raginsky (2017) provides bounds on the generalization error of a learning
algorithm in terms of the mutual information between the algorithm's output and
the training sample. In this work, we study the proposal, by Steinke and
Zakynthinou (2020), to reason about the generalization error of a learning
algorithm by introducing a super sample that contains the training sample as a
random subset and computing mutual information conditional on the super sample.
We first show that these new bounds based on the conditional mutual information
are tighter than those based on the unconditional mutual information. We then
introduce yet tighter bounds, building on the "individual sample" idea of Bu,
S. Zou, and Veeravalli (2019) and the "data dependent" ideas of Negrea et al.
(2019), using disintegrated mutual information. Finally, we apply these bounds
to the study of Langevin dynamics algorithm, showing that conditioning on the
super sample allows us to exploit information in the optimization trajectory to
obtain tighter bounds based on hypothesis tests.
| stat |
A Bayesian Approach to Linking Data Without Unique Identifiers | Existing file linkage methods may produce sub-optimal results because they
consider neither the interactions between different pairs of matched records
nor relationships between variables that are exclusive to one of the files. In
addition, many of the current methods fail to address the uncertainty in the
linkage, which may result in overly precise estimates of relationships between
variables that are exclusive to one of the files. Bayesian methods for record
linkage can reduce the bias in the estimation of scientific relationships of
interest and provide interval estimates that account for the uncertainty in the
linkage; however, implementation of these methods can often be complex and
computationally intensive. This article presents the GFS package for the R
programming language that utilizes a Bayesian approach for file linkage. The
linking procedure implemented in GFS samples from the joint posterior
distribution of model parameters and the linking permutations. The algorithm
approaches file linkage as a missing data problem and generates multiple linked
data sets. For computational efficiency, only the linkage permutations are
stored and multiple analyses are performed using each of the permutations
separately. This implementation reduces the computational complexity of the
linking process and the expertise required of researchers analyzing linked data
sets. We describe the algorithm implemented in the GFS package and its
statistical basis, and demonstrate its use on a sample data set.
| stat |
Likelihood-based missing data analysis in multivariate crossover trials | For gene expression data measured in a crossover trial, a multivariate
mixed-effects model seems to be most appropriate. Standard statistical
inference fails to provide reliable results when some responses are missing.
Particularly for crossover studies, missingness is a serious concern as the
trial requires a small number of participants. A Monte Carlo EM (MCEM) based
technique has been adopted to deal with this situation. Along with estimation,
a MCEM likelihood ratio test (LRTs) is developed for testing the fixed effects
in such a multivariate crossover model with missing data. Intensive simulation
studies have been carried out prior to the analysis of the gene expression
data.
| stat |
Root Cause Analysis in Lithium-Ion Battery Production with FMEA-Based
Large-Scale Bayesian Network | The production of lithium-ion battery cells is characterized by a high degree
of complexity due to numerous cause-effect relationships between process
characteristics. Knowledge about the multi-stage production is spread among
several experts, rendering tasks as failure analysis challenging. In this
paper, a new method is presented that includes expert knowledge acquisition in
production ramp-up by combining Failure Mode and Effects Analysis (FMEA) with a
Bayesian Network. Special algorithms are presented that help detect and resolve
inconsistencies between the expert-provided parameters which are bound to occur
when collecting knowledge from several process experts. We show the
effectiveness of this holistic method by building up a large scale,
cross-process Bayesian Failure Network in lithium-ion battery production and
its application for root cause analysis.
| stat |
Conformal prediction intervals for the individual treatment effect | We propose several prediction intervals procedures for the individual
treatment effect with either finite-sample or asymptotic coverage guarantee in
a non-parametric regression setting, where non-linear regression functions,
heteroskedasticity and non-Gaussianity are allowed. The construct the
prediction intervals we use the conformal method of Vovk et al. (2005). In
extensive simulations, we compare the coverage probability and interval length
of our prediction interval procedures. We demonstrate that complex learning
algorithms, such as neural networks, can lead to narrower prediction intervals
than simple algorithms, such as linear regression, if the sample size is large
enough.
| stat |
Supervised Principal Component Regression for Functional Response with
High Dimensional Predictors | We propose a supervised principal component regression method for relating
functional responses with high dimensional covariates. Unlike the conventional
principal component analysis, the proposed method builds on a newly defined
expected integrated residual sum of squares, which directly makes use of the
association between functional response and predictors. Minimizing the
integrated residual sum of squares gives the supervised principal components,
which is equivalent to solving a sequence of nonconvex generalized Rayleigh
quotient optimization problems and thus is computationally intractable. To
overcome this computational challenge, we reformulate the nonconvex
optimization problems into a simultaneous linear regression, with a sparse
penalty added to deal with high dimensional predictors. Theoretically, we show
that the reformulated regression problem recovers the same supervised principal
subspace under suitable conditions. Statistically, we establish non-asymptotic
error bounds for the proposed estimators. Numerical studies and an application
to the Human Connectome Project lend further support.
| stat |
Predicting class-imbalanced business risk using resampling,
regularization, and model ensembling algorithms | We aim at developing and improving the imbalanced business risk modeling via
jointly using proper evaluation criteria, resampling, cross-validation,
classifier regularization, and ensembling techniques. Area Under the Receiver
Operating Characteristic Curve (AUC of ROC) is used for model comparison based
on 10-fold cross validation. Two undersampling strategies including random
undersampling (RUS) and cluster centroid undersampling (CCUS), as well as two
oversampling methods including random oversampling (ROS) and Synthetic Minority
Oversampling Technique (SMOTE), are applied. Three highly interpretable
classifiers, including logistic regression without regularization (LR),
L1-regularized LR (L1LR), and decision tree (DT) are implemented. Two
ensembling techniques, including Bagging and Boosting, are applied on the DT
classifier for further model improvement. The results show that, Boosting on DT
by using the oversampled data containing 50% positives via SMOTE is the optimal
model and it can achieve AUC, recall, and F1 score valued 0.8633, 0.9260, and
0.8907, respectively.
| stat |
Parametric quantile regression models for fitting double bounded
response with application to COVID-19 mortality rate data | In this paper, we develop two fully parametric quantile regression models,
based on power Johnson SB distribution Cancho et al. (2020), for modeling unit
interval response at different quantiles. In particular, the conditional
distribution is modelled by the power Johnson SB distribution. The maximum
likelihood method is employed to estimate the model parameters. Simulation
studies are conducted to evaluate the performance of the maximum likelihood
estimators in finite samples. Furthermore, we discuss residuals and influence
diagnostic tools. The effectiveness of our proposals is illustrated with two
data set given by the mortality rate of COVID-19 in different countries.
| stat |
Late 19th-Century Navigational Uncertainties and Their Influence on Sea
Surface Temperature Estimates | Accurate estimates of historical changes in sea surface temperatures (SSTs)
and their uncertainties are important for documenting and understanding
historical changes in climate. A source of uncertainty that has not previously
been quantified in historical SST estimates stems from position errors. A
Bayesian inference framework is proposed for quantifying errors in reported
positions and their implications on SST estimates. The analysis framework is
applied to data from the International Comprehensive Ocean-Atmosphere Data Set
(ICOADS3.0) in 1885, a time when astronomical and chronometer estimation of
position was common, but predating the use of radio signals. Focus is upon a
subset of 943 ship tracks from ICOADS3.0 that report their position every two
hours to a precision of 0.01{\deg} longitude and latitude. These data are
interpreted as positions determined by dead reckoning that are periodically
updated by celestial correction techniques. The posterior medians of
uncertainties in celestial correction are 33.1 km (0.30{\deg} on the equator)
in longitude and 24.4 km (0.22{\deg}) in latitude, respectively. The posterior
medians of two-hourly dead reckoning uncertainties are 19.2% for ship speed and
13.2{\deg} for ship heading, leading to random position uncertainties with
median 0.18{\deg} (20 km on the equator) in longitude and 0.15{\deg} (17 km) in
latitude. Reported ship tracks also contain systematic position uncertainties
relating to precursor dead-reckoning positions not being updated after
obtaining celestial position estimates, indicating that more accurate positions
can be provided for SST observations. Finally, we translate position errors
into SST uncertainties by sampling an ensemble of SSTs from the Multi-scale
Ultra-high resolution Sea Surface Temperature (MURSST) data set.
| stat |
Towards a mathematical theory of trajectory inference | We devise a theoretical framework and a numerical method to infer
trajectories of a stochastic process from snapshots of its temporal marginals.
This problem arises in the analysis of single cell RNA-sequencing data, which
provide high dimensional measurements of cell states but cannot track the
trajectories of the cells over time. We prove that for a class of stochastic
processes it is possible to recover the ground truth trajectories from limited
samples of the temporal marginals at each time-point, and provide an efficient
algorithm to do so in practice. The method we develop, Global Waddington-OT
(gWOT), boils down to a smooth convex optimization problem posed globally over
all time-points involving entropy-regularized optimal transport. We demonstrate
that this problem can be solved efficiently in practice and yields good
reconstructions, as we show on several synthetic and real datasets.
| stat |
Sparse and Functional Principal Components Analysis | Regularized variants of Principal Components Analysis, especially Sparse PCA
and Functional PCA, are among the most useful tools for the analysis of complex
high-dimensional data. Many examples of massive data, have both sparse and
functional (smooth) aspects and may benefit from a regularization scheme that
can capture both forms of structure. For example, in neuro-imaging data, the
brain's response to a stimulus may be restricted to a discrete region of
activation (spatial sparsity), while exhibiting a smooth response within that
region. We propose a unified approach to regularized PCA which can induce both
sparsity and smoothness in both the row and column principal components. Our
framework generalizes much of the previous literature, with sparse, functional,
two-way sparse, and two-way functional PCA all being special cases of our
approach. Our method permits flexible combinations of sparsity and smoothness
that lead to improvements in feature selection and signal recovery, as well as
more interpretable PCA factors. We demonstrate the efficacy of our method on
simulated data and a neuroimaging example on EEG data.
| stat |
Dynamical analysis in a self-regulated system undergoing multiple
excitations: first order differential equation approach | This article proposes a dynamical system modeling approach for the analysis
of longitudinal data of self-regulated systems experiencing multiple
excitations. The aim of such an approach is to focus on the evolution of a
signal (e.g., heart rate) before, during, and after excitations taking the
system out of its equilibrium (e.g., physical effort during cardiac stress
testing). Dynamical modeling can be applied to a broad range of outcomes such
as physiological processes in medicine and psychosocial processes in social
sciences, and it allows to extract simple characteristics of the signal
studied. The model we propose is based on a first order linear differential
equation defined by three main parameters corresponding to the initial
equilibrium value, the dynamic characteristic time, and the reaction to the
excitation. In this paper, several estimation procedures for this model are
considered and tested in a simulation study, that clarifies under which
conditions accurate estimates are provided. Finally, applications of this model
are illustrated using cardiology data recorded during effort tests.
| stat |