title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Moving from two- to multi-way interactions among binary risk factors on
the additive scale | Many studies have focused on investigating deviations from additive
interaction of two dichotomous risk factors on a binary outcome. There is,
however, a gap in the literature with respect to interactions on the additive
scale of >2 risk factors. In this paper, we present an approach for examining
deviations from additive interaction among three on more binary exposures. The
relative excess risk due to interaction (RERI) is used as measure of additive
interaction. First, we concentrate on three risk factors - we propose to
decompose the total RERI to: the RERI owned to the joint presence of all 3 risk
factors and the RERI of any two risk factors, given that the third is absent.
We then extend this approach, to >3 binary risk factors. For illustration, we
use a sample from data from the Greek EPIC cohort and we investigate the
association with overall mortality of Mediterranean diet, body mass index
(BMI), and, smoking. Our formulae enable better interpretability of any
evidence for deviations from additivity owned to more than two risk factors and
provide simple ways of communicating such results from a public health
perspective by attributing any excess relative risk to specific combinations of
these factors.
| stat |
Weighted-Graph-Based Change Point Detection | We consider the detection and localization of change points in the
distribution of an offline sequence of observations. Based on a nonparametric
framework that uses a similarity graph among observations, we propose new test
statistics when at most one change point occurs and generalize them to multiple
change points settings. The proposed statistics leverage edge weight
information in the graphs, exhibiting substantial improvements in testing power
and localization accuracy in simulations. We derive the null limiting
distribution, provide accurate analytic approximations to control type I error,
and establish theoretical guarantees on the power consistency under contiguous
alternatives for the one change point setting, as well as the minimax
localization rate. In the multiple change points setting, the asymptotic
correctness of the number and location of change points are also guaranteed.
The methods are illustrated on the MIT proximity network data.
| stat |
Sparse Identification and Estimation of Large-Scale Vector
AutoRegressive Moving Averages | The Vector AutoRegressive Moving Average (VARMA) model is fundamental to the
theory of multivariate time series; however, in practice, identifiability
issues have led many authors to abandon VARMA modeling in favor of the simpler
Vector AutoRegressive (VAR) model. Such a practice is unfortunate since even
very simple VARMA models can have quite complicated VAR representations. We
narrow this gap with a new optimization-based approach to VARMA identification
that is built upon the principle of parsimony. Among all equivalent
data-generating models, we seek the parameterization that is "simplest" in a
certain sense. A user-specified strongly convex penalty is used to measure
model simplicity, and that same penalty is then used to define an estimator
that can be efficiently computed. We show that our estimator converges to a
parsimonious element in the set of all equivalent data-generating models, in a
double asymptotic regime where the number of component time series is allowed
to grow with sample size. Further, we derive non-asymptotic upper bounds on the
estimation error of our method relative to our specially identified target.
Novel theoretical machinery includes non-asymptotic analysis of infinite-order
VAR, elastic net estimation under a singular covariance structure of
regressors, and new concentration inequalities for quadratic forms of random
variables from Gaussian time series. We illustrate the competitive performance
of our methods in simulation and several application domains, including
macro-economic forecasting, demand forecasting, and volatility forecasting.
| stat |
PAC-Bayes unleashed: generalisation bounds with unbounded losses | We present new PAC-Bayesian generalisation bounds for learning problems with
unbounded loss functions. This extends the relevance and applicability of the
PAC-Bayes learning framework, where most of the existing literature focuses on
supervised learning problems with a bounded loss function (typically assumed to
take values in the interval [0;1]). In order to relax this assumption, we
propose a new notion called HYPE (standing for \emph{HYPothesis-dependent
rangE}), which effectively allows the range of the loss to depend on each
predictor. Based on this new notion we derive a novel PAC-Bayesian
generalisation bound for unbounded loss functions, and we instantiate it on a
linear regression problem. To make our theory usable by the largest audience
possible, we include discussions on actual computation, practicality and
limitations of our assumptions.
| stat |
Quantifying Age and Model Uncertainties in Paleoclimate Data and
Dynamical Climate Models with a Joint Inferential Analysis | A major goal in paleoclimate science is to reconstruct historical climates
using proxies for climate variables such as those observed in sediment cores,
and in the process learn about climate dynamics. This is hampered by
uncertainties in how sediment core depths relate to ages, how proxy quantities
relate to climate variables, how climate models are specified, and the values
of parameters in climate models. Quantifying these uncertainties is key in
drawing well founded conclusions. Analyses are often performed in separate
stages with, for example, a sediment core's depth-age relation being estimated
as stage one, then fed as an input to calibrate climate models as stage two.
Here, we show that such "multi-stage" approaches can lead to misleading
conclusions. We develop a joint inferential approach for climate
reconstruction, model calibration, and age model estimation. We focus on the
glacial-interglacial cycle over the past 780 kyr, analysing two sediment cores
that span this range. Our age estimates are largely in agreement with previous
studies, but provides the full joint specification of all uncertainties,
estimation of model parameters, and the model evidence. By sampling plausible
chronologies from the posterior distribution, we demonstrate that downstream
scientific conclusions can differ greatly both between different sampled
chronologies, and in comparison with conclusions obtained in the complete joint
inferential analysis. We conclude that multi-stage analyses are insufficient
when dealing with uncertainty, and that to draw sound conclusions the full
joint inferential analysis must be performed.
| stat |
Ensemble Neural Networks (ENN): A gradient-free stochastic method | In this study, an efficient stochastic gradient-free method, the ensemble
neural networks (ENN), is developed. In the ENN, the optimization process
relies on covariance matrices rather than derivatives. The covariance matrices
are calculated by the ensemble randomized maximum likelihood algorithm (EnRML),
which is an inverse modeling method. The ENN is able to simultaneously provide
estimations and perform uncertainty quantification since it is built under the
Bayesian framework. The ENN is also robust to small training data size because
the ensemble of stochastic realizations essentially enlarges the training
dataset. This constitutes a desirable characteristic, especially for real-world
engineering applications. In addition, the ENN does not require the calculation
of gradients, which enables the use of complicated neuron models and loss
functions in neural networks. We experimentally demonstrate benefits of the
proposed model, in particular showing that the ENN performs much better than
the traditional Bayesian neural networks (BNN). The EnRML in ENN is a
substitution of gradient-based optimization algorithms, which means that it can
be directly combined with the feed-forward process in other existing (deep)
neural networks, such as convolutional neural networks (CNN) and recurrent
neural networks (RNN), broadening future applications of the ENN.
| stat |
Reduced-Rank Tensor-on-Tensor Regression and Tensor-variate Analysis of
Variance | Fitting regression models with many multivariate responses and covariates can
be challenging, but such responses and covariates sometimes have tensor-variate
structure. We extend the classical multivariate regression model to exploit
such structure in two ways: first, we impose four types of low-rank tensor
formats on the regression coefficients. Second, we model the errors using the
tensor-variate normal distribution that imposes a Kronecker separable format on
the covariance matrix. We obtain maximum likelihood estimators via
block-relaxation algorithms and derive their asymptotic distributions. Our
regression framework enables us to formulate tensor-variate analysis of
variance (TANOVA) methodology. Application of our methodology in a one-way
TANOVA layout enables us to identify cerebral regions significantly associated
with the interaction of suicide attempters or non-attemptor ideators and
positive-, negative- or death-connoting words. A separate application performs
three-way TANOVA on the Labeled Faces in the Wild image database to distinguish
facial characteristics related to ethnic origin, age group and gender.
| stat |
Differential privacy and noisy confidentiality concepts for European
population statistics | The paper aims to give an overview of various approaches to statistical
disclosure control based on random noise that are currently being discussed for
official population statistics and censuses. A particular focus is on a
stringent delineation between different concepts influencing the discussion: we
separate clearly between risk measures, noise distributions and output
mechanisms - putting these concepts into scope and into relation with each
other.
After recapitulating differential privacy as a risk measure, the paper also
remarks on utility and risk aspects of some specific output mechanisms and
parameter setups, with special attention on static outputs that are rather
typical in official population statistics. In particular, it is argued that
unbounded noise distributions, such as plain Laplace, may jeopardise key unique
census features without a clear need from a risk perspective. On the other
hand, bounded noise distributions, such as the truncated Laplace or the cell
key method, can be set up to keep unique census features while controlling
disclosure risks in census-like outputs.
Finally, the paper analyses some typical attack scenarios to constrain
generic noise parameter ranges that suggest a good risk/utility compromise for
the 2021 EU census output scenario. The analysis also shows that strictly
differentially private mechanisms would be severely constrained in this
scenario.
| stat |
Functional Principal Component Analysis of Cointegrated Functional Time
Series | Functional principal component analysis (FPCA) has played an important role
in the development of functional time series analysis. This paper investigates
how FPCA can be used to analyze cointegrated functional time series and
proposes a modification of FPCA as a novel statistical tool. Our modified FPCA
not only provides an asymptotically more efficient estimator of the
cointegrating vectors, but also leads to novel FPCA-based tests for examining
some essential properties of cointegrated functional time series. As an
empirical illustration, our methodology is applied to two empirical examples:
age-specific employment rates and earning densities.
| stat |
WOMBAT: A fully Bayesian global flux-inversion framework | WOMBAT (the WOllongong Methodology for Bayesian Assimilation of Trace-gases)
is a fully Bayesian hierarchical statistical framework for flux inversion of
trace gases from flask, in situ, and remotely sensed data. WOMBAT extends the
conventional Bayesian-synthesis framework through the consideration of a
correlated error term, the capacity for online bias correction, and the
provision of uncertainty quantification on all unknowns that appear in the
Bayesian statistical model. We show, in an observing system simulation
experiment (OSSE), that these extensions are crucial when the data are indeed
biased and have errors that are correlated. Using the GEOS-Chem atmospheric
transport model, we show that WOMBAT is able to obtain posterior means and
uncertainties on non-fossil-fuel CO$_2$ fluxes from Orbiting Carbon
Observatory-2 (OCO-2) data that are comparable to those from the Model
Intercomparison Project (MIP) reported in Crowell et al. (2019, Atmos. Chem.
Phys., vol. 19). We also find that our predictions of out-of-sample retrievals
from the Total Column Carbon Observing Network are, for the most part, more
accurate than those made by the MIP participants. Subsequent versions of the
OCO-2 datasets will be ingested into WOMBAT as they become available.
| stat |
Fast Bayesian Inference in Nonparametric Double Additive Location-Scale
Models With Right- and Interval-Censored Data | Penalized B-splines are routinely used in additive models to describe smooth
changes in a response with quantitative covariates. It is typically done
through the conditional mean in the exponential family using generalized
additive models with an indirect impact on other conditional moments. Another
common strategy consists in focussing on several low-order conditional moments,
leaving the complete conditional distribution unspecified. Alternatively, a
multi-parameter distribution could be assumed for the response with several of
its parameters jointly regressed on covariates using additive expressions.
Our work can be connected to the latter proposal for a right- or
interval-censored continuous response with a highly flexible and smooth
nonparametric density. We focus on location-scale models with additive terms in
the conditional mean and standard deviation. Starting from recent results in
the Bayesian framework, we propose a quickly converging algorithm to select
penalty parameters from their marginal posteriors. It relies on Laplace
approximations to the conditional posterior of the spline parameters.
Simulations suggest that the so-obtained estimators own excellent frequentist
properties and increase efficiency as compared to approaches with a working
Gaussian hypothesis. We illustrate the methodology with the analysis of
imprecisely measured income data.
| stat |
A new robust class of skew elliptical distributions | A new robust class of multivariate skew distributions is introduced.
Practical aspects such as parameter estimation method of the proposed class are
discussed, we show that the proposed class can be fitted under a reasonable
time frame. Our study shows that the class of distributions is capable to model
multivariate skewness structure and does not suffer from the curse of
dimensionality as heavily as other distributions of similar complexity do, such
as the class of canonical skew distributions. We also derive a nested form of
the proposed class which appears to be the most flexible class of multivariate
skew distributions in literature that has a closed-form density function.
Numerical examples on two data sets, i) a data set containing daily river flow
data recorded in the UK; and ii) a data set containing biomedical variables of
athletes collected by the Australian Institute of Sports (AIS), are
demonstrated. These examples further support the practicality of the proposed
class on moderate dimensional data sets.
| stat |
Comparing Semi-Parametric Model Learning Algorithms for Dynamic Model
Estimation in Robotics | Physical modeling of robotic system behavior is the foundation for
controlling many robotic mechanisms to a satisfactory degree. Mechanisms are
also typically designed in a way that good model accuracy can be achieved with
relatively simple models and model identification strategies. If the modeling
accuracy using physically based models is not enough or too complex, model-free
methods based on machine learning techniques can help. Of particular interest
to us was therefore the question to what degree semi-parametric modeling
techniques, meaning combinations of physical models with machine learning,
increase the modeling accuracy of inverse dynamics models which are typically
used in robot control. To this end, we evaluated semi-parametric Gaussian
process regression and a novel model-based neural network architecture, and
compared their modeling accuracy to a series of naive semi-parametric,
parametric-only and non-parametric-only regression methods. The comparison has
been carried out on three test scenarios, one involving a real test-bed and two
involving simulated scenarios, with the most complex scenario targeting the
modeling a simulated robot's inverse dynamics model. We found that in all but
one case, semi-parametric Gaussian process regression yields the most accurate
models, also with little tuning required for the training procedure.
| stat |
A General Framework of Online Updating Variable Selection for
Generalized Linear Models with Streaming Datasets | In the research field of big data, one of important issues is how to recover
the sequentially changing sets of true features when the data sets arrive
sequentially. The paper presents a general framework for online updating
variable selection and parameter estimation in generalized linear models with
streaming datasets. This is a type of online updating penalty likelihoods with
differentiable or non-differentiable penalty function. The online updating
coordinate descent algorithm is proposed to solve the online updating
optimization problem. Moreover, a tuning parameter selection is suggested in an
online updating way. The selection and estimation consistencies, and the oracle
property are established, theoretically. Our methods are further examined and
illustrated by various numerical examples from both simulation experiments and
a real data analysis.
| stat |
Stochastic Runge-Kutta Accelerates Langevin Monte Carlo and Beyond | Sampling with Markov chain Monte Carlo methods often amounts to discretizing
some continuous-time dynamics with numerical integration. In this paper, we
establish the convergence rate of sampling algorithms obtained by discretizing
smooth It\^o diffusions exhibiting fast Wasserstein-$2$ contraction, based on
local deviation properties of the integration scheme. In particular, we study a
sampling algorithm constructed by discretizing the overdamped Langevin
diffusion with the method of stochastic Runge-Kutta. For strongly convex
potentials that are smooth up to a certain order, its iterates converge to the
target distribution in $2$-Wasserstein distance in
$\tilde{\mathcal{O}}(d\epsilon^{-2/3})$ iterations. This improves upon the
best-known rate for strongly log-concave sampling based on the overdamped
Langevin equation using only the gradient oracle without adjustment. In
addition, we extend our analysis of stochastic Runge-Kutta methods to uniformly
dissipative diffusions with possibly non-convex potentials and show they
achieve better rates compared to the Euler-Maruyama scheme in terms of the
dependence on tolerance $\epsilon$. Numerical studies show that these
algorithms lead to better stability and lower asymptotic errors.
| stat |
Reference Product Search | For a product of interest, we propose a search method to surface a set of
reference products. The reference products can be used as candidates to support
downstream modeling tasks and business applications. The search method consists
of product representation learning and fingerprint-type vector searching. The
product catalog information is transformed into a high-quality embedding of low
dimensions via a novel attention auto-encoder neural network, and the embedding
is further coupled with a binary encoding vector for fast retrieval. We conduct
extensive experiments to evaluate the proposed method, and compare it with peer
services to demonstrate its advantage in terms of search return rate and
precision.
| stat |
Pointer Graph Networks | Graph neural networks (GNNs) are typically applied to static graphs that are
assumed to be known upfront. This static input structure is often informed
purely by insight of the machine learning practitioner, and might not be
optimal for the actual task the GNN is solving. In absence of reliable domain
expertise, one might resort to inferring the latent graph structure, which is
often difficult due to the vast search space of possible graphs. Here we
introduce Pointer Graph Networks (PGNs) which augment sets or graphs with
additional inferred edges for improved model generalisation ability. PGNs allow
each node to dynamically point to another node, followed by message passing
over these pointers. The sparsity of this adaptable graph structure makes
learning tractable while still being sufficiently expressive to simulate
complex algorithms. Critically, the pointing mechanism is directly supervised
to model long-term sequences of operations on classical data structures,
incorporating useful structural inductive biases from theoretical computer
science. Qualitatively, we demonstrate that PGNs can learn parallelisable
variants of pointer-based data structures, namely disjoint set unions and
link/cut trees. PGNs generalise out-of-distribution to 5x larger test inputs on
dynamic graph connectivity tasks, outperforming unrestricted GNNs and Deep
Sets.
| stat |
Competing risks joint models using R-INLA | The methodological advancements made in the field of joint models are
numerous. None the less, the case of competing risks joint models have largely
been neglected, especially from a practitioner's point of view. In the relevant
works on competing risks joint models, the assumptions of Gaussian linear
longitudinal series and proportional cause-specific hazard functions, amongst
others, have remained unchallenged. In this paper, we provide a framework based
on R-INLA to apply competing risks joint models in a unifying way such that
non-Gaussian longitudinal data, spatial structures, time dependent splines and
various latent association structures, to mention a few, are all embraced in
our approach. Our motivation stems from the SANAD trial which exhibits
non-linear longitudinal trajectories and competing risks for failure of
treatment. We also present a discrete competing risks joint model for
longitudinal count data as well as a spatial competing risks joint model, as
specific examples.
| stat |
Model Distillation for Revenue Optimization: Interpretable Personalized
Pricing | Data-driven pricing strategies are becoming increasingly common, where
customers are offered a personalized price based on features that are
predictive of their valuation of a product. It is desirable to have this
pricing policy be simple and interpretable, so it can be verified, checked for
fairness, and easily implemented. However, efforts to incorporate machine
learning into a pricing framework often lead to complex pricing policies which
are not interpretable, resulting in mixed results in practice. We present a
customized, prescriptive tree-based algorithm that distills knowledge from a
complex black box machine learning algorithm, segments customers with similar
valuations and prescribes prices in such a way that maximizes revenue while
maintaining interpretability. We quantify the regret of a resulting policy and
demonstrate its efficacy in applications with both synthetic and real-world
datasets.
| stat |
Directional quantile classifiers | We introduce classifiers based on directional quantiles. We derive
theoretical results for selecting optimal quantile levels given a direction,
and, conversely, an optimal direction given a quantile level. We also show that
the misclassification rate is infinitesimal if population distributions differ
by at most a location shift and if the number of directions is allowed to
diverge at the same rate of the problem's dimension. We illustrate the
satisfactory performance of our proposed classifiers in both small and high
dimensional settings via a simulation study and a real data example. The code
implementing the proposed methods is publicly available in the R package
Qtools.
| stat |
A Causal Machine Learning Framework for Predicting Preventable Hospital
Readmissions | Clinical predictive algorithms are increasingly being used to form the basis
for optimal treatment policies--that is, to enable interventions to be targeted
to the patients who will presumably benefit most. Despite taking advantage of
recent advances in supervised machine learning, these algorithms remain, in a
sense, blunt instruments--often being developed and deployed without a full
accounting of the causal aspects of the prediction problems they are intended
to solve. Indeed, in many settings, including among patients at risk of
readmission, the riskiest patients may derive less benefit from a preventative
intervention compared to those at lower risk. Moreover, targeting an
intervention to a population, rather than limiting it to a small group of
high-risk patients, may lead to far greater overall utility if the patients
with the most modifiable (or preventable) outcomes across the population could
be identified. Based on these insights, we introduce a causal machine learning
framework that decouples this prediction problem into causal and predictive
parts, which clearly delineates the complementary roles of causal inference and
prediction in this problem. We estimate treatment effects using causal forests,
and characterize treatment effect heterogeneity across levels of predicted risk
using these estimates. Furthermore, we show how these effect estimates could be
used in concert with the modeled "payoffs" associated with successful
prevention of individual readmissions to maximize overall utility. Based on
data taken from before and after the implementation of a readmissions
prevention intervention at Kaiser Permanente Northern California, our results
suggest that nearly four times as many readmissions could be prevented annually
with this approach compared to targeting this intervention using predicted
risk.
| stat |
A powerful MAF-neutral allele-based test for case-control association
studies | In a case-control study aimed at locating autosomal disease variants for a
disease of interest, association between markers and the disease status is
often tested by comparing the marker minor allele frequencies (MAFs) between
cases and controls. For most common allele-based tests the statistical power is
highly dependent on the actual values of these MAFs, where associated markers
with low MAFs have less power to be detected compared to associated markers
with high MAFs. Therefore, the popular strategy of selecting markers for
follow-up studies based primarily on their p-values is likely to preferentially
select markers with high MAFs. We propose a new test which does not favor
markers with high MAFs and improves the power for markers with low to moderate
MAFs without sacrificing performance for markers with high MAFs and is
therefore superior to most existing tests in this regard. An explicit formula
for the asymptotic power function of the proposed test is derived
theoretically, which allows for fast and easy computation of the corresponding
p-values. The performance of the proposed test is compared with several
existing tests both in the asymptotic and the finite sample size settings.
| stat |
Sample Size Re-estimation Design in Phase II Dose Finding Study with
Multiple Dose Groups: Frequentist and Bayesian Methods | Unblinded sample size re-estimation (SSR) is often planned in a clinical
trial when there is large uncertainty about the true treatment effect. For
Proof-of Concept (PoC) in a Phase II dose finding study, contrast test can be
adopted to leverage information from all treatment groups. In this article, we
propose two-stage SSR designs using frequentist conditional power and Bayesian
posterior predictive power for both single and multiple contrast tests. The
Bayesian SSR can be implemented under a wide range of prior settings to
incorporate different prior knowledge. Taking the adaptivity into account, all
type I errors of final analysis in this paper are rigorously protected.
Simulation studies are carried out to demonstrate the advantages of unblinded
SSR in multi-arm trials.
| stat |
Causal inference of brain connectivity from fMRI with $\psi$-Learning
Incorporated Linear non-Gaussian Acyclic Model ($\psi$-LiNGAM) | Functional connectivity (FC) has become a primary means of understanding
brain functions by identifying brain network interactions and, ultimately, how
those interactions produce cognitions. A popular definition of FC is by
statistical associations between measured brain regions. However, this could be
problematic since the associations can only provide spatial connections but not
causal interactions among regions of interests. Hence, it is necessary to study
their causal relationship. Directed acyclic graph (DAG) models have been
applied in recent FC studies but often encountered problems such as limited
sample sizes and large number of variables (namely high-dimensional problems),
which lead to both computational difficulty and convergence issues. As a
result, the use of DAG models is problematic, where the identification of DAG
models in general is nondeterministic polynomial time hard (NP-hard). To this
end, we propose a $\psi$-learning incorporated linear non-Gaussian acyclic
model ($\psi$-LiNGAM). We use the association model ($\psi$-learning) to
facilitate causal inferences and the model works well especially for
high-dimensional cases. Our simulation results demonstrate that the proposed
method is more robust and accurate than several existing ones in detecting
graph structure and direction. We then applied it to the resting state fMRI
(rsfMRI) data obtained from the publicly available Philadelphia
Neurodevelopmental Cohort (PNC) to study the cognitive variance, which includes
855 individuals aged 8-22 years. Therein, we have identified three types of hub
structure: the in-hub, out-hub and sum-hub, which correspond to the centers of
receiving, sending and relaying information, respectively. We also detected 16
most important pairs of causal flows. Several of the results have been verified
to be biologically significant.
| stat |
Local dimension reduction of summary statistics for likelihood-free
inference | Approximate Bayesian computation (ABC) and other likelihood-free inference
methods have gained popularity in the last decade, as they allow rigorous
statistical inference for complex models without analytically tractable
likelihood functions. A key component for accurate inference with ABC is the
choice of summary statistics, which summarize the information in the data, but
at the same time should be low-dimensional for efficiency. Several dimension
reduction techniques have been introduced to automatically construct
informative and low-dimensional summaries from a possibly large pool of
candidate summaries. Projection-based methods, which are based on learning
simple functional relationships from the summaries to parameters, are widely
used and usually perform well, but might fail when the assumptions behind the
transformation are not satisfied. We introduce a localization strategy for any
projection-based dimension reduction method, in which the transformation is
estimated in the neighborhood of the observed data instead of the whole space.
Localization strategies have been suggested before, but the performance of the
transformed summaries outside the local neighborhood has not been guaranteed.
In our localization approach the transformation is validated and optimized over
validation datasets, ensuring reliable performance. We demonstrate the
improvement in the estimation accuracy for localized versions of linear
regression and partial least squares, for three different models of varying
complexity.
| stat |
Generating Random Samples from Non-Identical Truncated Order Statistics | We provide an efficient algorithm to generate random samples from the bounded
kth order statistic in a sample of independent, but not necessarily identically
distributed, random variables. The bounds can be upper or lower bounds and need
only hold on the kth order statistic. Furthermore, we require access to the
inverse CDF for each statistic in the ordered sample. The algorithm is slightly
slower than rejection sampling when the density of the bounded statistic is
large, however, it is significantly faster when the bounded density becomes
sparse. We provide a practical example and a simulation that shows the
superiority of this method for sparse regions arising from tight boundary
conditions and/or over regions of low probability density.
| stat |
Empirical Bayes methods for monitoring health care quality | The paper discusses empirical Bayes methodology for repeated quality
comparisons of health care institutions using data from the Dutch VOKS study
that annually monitors the relative performance and quality of nearly all Dutch
gynecological centres with respect to different aspects of the childbirths
taking place in these centres. This paper can be seen as an extension of the
pioneering work of Thomas, Longford and Rolph and Goldstein and Spiegelhalter.
First of all, this paper introduces a new simple crude estimate of the centre
effect in a logistic regression setting. Next, a simple estimate of the
expected percentile of a centre given all data and a measure of rankability of
the centres based on the expected percentiles are presented. Finally, the
temporal dimension is explored and methods are discussed to predict next years
performance.
| stat |
Analyzing Factors Associated with Fatal Road Crashes: A Machine Learning
Approach | Road traffic injury accounts for a substantial human and economic burden
globally. Understanding risk factors contributing to fatal injuries is of
paramount importance. In this study, we proposed a model that adopts a hybrid
ensemble machine learning classifier structured from sequential minimal
optimization and decision trees to identify risk factors contributing to fatal
road injuries. The model was constructed, trained, tested, and validated using
the Lebanese Road Accidents Platform (LRAP) database of 8482 road crash
incidents, with fatality occurrence as the outcome variable. A sensitivity
analysis was conducted to examine the influence of multiple factors on fatality
occurrence. Seven out of the nine selected independent variables were
significantly associated with fatality occurrence, namely, crash type, injury
severity, spatial cluster-ID, and crash time (hour). Evidence gained from the
model data analysis will be adopted by policymakers and key stakeholders to
gain insights into major contributing factors associated with fatal road
crashes and to translate knowledge into safety programs and enhanced road
policies.
| stat |
Scalable Gaussian Process Classification with Additive Noise for Various
Likelihoods | Gaussian process classification (GPC) provides a flexible and powerful
statistical framework describing joint distributions over function space.
Conventional GPCs however suffer from (i) poor scalability for big data due to
the full kernel matrix, and (ii) intractable inference due to the non-Gaussian
likelihoods. Hence, various scalable GPCs have been proposed through (i) the
sparse approximation built upon a small inducing set to reduce the time
complexity; and (ii) the approximate inference to derive analytical evidence
lower bound (ELBO). However, these scalable GPCs equipped with analytical ELBO
are limited to specific likelihoods or additional assumptions. In this work, we
present a unifying framework which accommodates scalable GPCs using various
likelihoods. Analogous to GP regression (GPR), we introduce additive noises to
augment the probability space for (i) the GPCs with step, (multinomial) probit
and logit likelihoods via the internal variables; and particularly, (ii) the
GPC using softmax likelihood via the noise variables themselves. This leads to
unified scalable GPCs with analytical ELBO by using variational inference.
Empirically, our GPCs showcase better results than state-of-the-art scalable
GPCs for extensive binary/multi-class classification tasks with up to two
million data points.
| stat |
Cross-validation: what does it estimate and how well does it do it? | Cross-validation is a widely-used technique to estimate prediction error, but
its behavior is complex and not fully understood. Ideally, one would like to
think that cross-validation estimates the prediction error for the model at
hand, fit to the training data. We prove that this is not the case for the
linear model fit by ordinary least squares; rather it estimates the average
prediction error of models fit on other unseen training sets drawn from the
same population. We further show that this phenomenon occurs for most popular
estimates of prediction error, including data splitting, bootstrapping, and
Mallow's Cp. Next, the standard confidence intervals for prediction error
derived from cross-validation may have coverage far below the desired level.
Because each data point is used for both training and testing, there are
correlations among the measured accuracies for each fold, and so the usual
estimate of variance is too small. We introduce a nested cross-validation
scheme to estimate this variance more accurately, and show empirically that
this modification leads to intervals with approximately correct coverage in
many examples where traditional cross-validation intervals fail. Lastly, our
analysis also shows that when producing confidence intervals for prediction
accuracy with simple data splitting, one should not re-fit the model on the
combined data, since this invalidates the confidence intervals.
| stat |
Discrete Action On-Policy Learning with Action-Value Critic | Reinforcement learning (RL) in discrete action space is ubiquitous in
real-world applications, but its complexity grows exponentially with the
action-space dimension, making it challenging to apply existing on-policy
gradient based deep RL algorithms efficiently. To effectively operate in
multidimensional discrete action spaces, we construct a critic to estimate
action-value functions, apply it on correlated actions, and combine these
critic estimated action values to control the variance of gradient estimation.
We follow rigorous statistical analysis to design how to generate and combine
these correlated actions, and how to sparsify the gradients by shutting down
the contributions from certain dimensions. These efforts result in a new
discrete action on-policy RL algorithm that empirically outperforms related
on-policy algorithms relying on variance control techniques. We demonstrate
these properties on OpenAI Gym benchmark tasks, and illustrate how discretizing
the action space could benefit the exploration phase and hence facilitate
convergence to a better local optimal solution thanks to the flexibility of
discrete policy.
| stat |
Machine Learning for Scent: Learning Generalizable Perceptual
Representations of Small Molecules | Predicting the relationship between a molecule's structure and its odor
remains a difficult, decades-old task. This problem, termed quantitative
structure-odor relationship (QSOR) modeling, is an important challenge in
chemistry, impacting human nutrition, manufacture of synthetic fragrance, the
environment, and sensory neuroscience. We propose the use of graph neural
networks for QSOR, and show they significantly out-perform prior methods on a
novel data set labeled by olfactory experts. Additional analysis shows that the
learned embeddings from graph neural networks capture a meaningful odor space
representation of the underlying relationship between structure and odor, as
demonstrated by strong performance on two challenging transfer learning tasks.
Machine learning has already had a large impact on the senses of sight and
sound. Based on these early results with graph neural networks for molecular
properties, we hope machine learning can eventually do for olfaction what it
has already done for vision and hearing.
| stat |
Improved Gradient-Based Optimization Over Discrete Distributions | In many applications we seek to maximize an expectation with respect to a
distribution over discrete variables. Estimating gradients of such objectives
with respect to the distribution parameters is a challenging problem. We
analyze existing solutions including finite-difference (FD) estimators and
continuous relaxation (CR) estimators in terms of bias and variance. We show
that the commonly used Gumbel-Softmax estimator is biased and propose a simple
method to reduce it. We also derive a simpler piece-wise linear continuous
relaxation that also possesses reduced bias. We demonstrate empirically that
reduced bias leads to a better performance in variational inference and on
binary optimization tasks.
| stat |
Traffic Flows Analysis in High-Speed Computer Networks Using Time Series | This article explores the required amount of time series points from a
high-speed traffic network to accurately estimate the Hurst exponent. The
methodology consists in designing an experiment using estimators that are
applied to time series, followed by addressing the minimum amount of points
required to obtain accurate estimates of the Hurst exponent in real-time. The
methodology addresses the exhaustive analysis of the Hurst exponent considering
bias behavior, standard deviation, mean square error, and convergence using
fractional gaussian noise signals with stationary increases. Our results show
that the Whittle estimator successfully estimates the Hurst exponent in series
with few points. Based on the results obtained, a minimum length for the time
series is empirically proposed. Finally, to validate the results, the
methodology is applied to real traffic captures in a high-speed network based
on the IEEE 802.3ab standard.
| stat |
Autoencoding Binary Classifiers for Supervised Anomaly Detection | We propose the Autoencoding Binary Classifiers (ABC), a novel supervised
anomaly detector based on the Autoencoder (AE). There are two main approaches
in anomaly detection: supervised and unsupervised. The supervised approach
accurately detects the known anomalies included in training data, but it cannot
detect the unknown anomalies. Meanwhile, the unsupervised approach can detect
both known and unknown anomalies that are located away from normal data points.
However, it does not detect known anomalies as accurately as the supervised
approach. Furthermore, even if we have labeled normal data points and
anomalies, the unsupervised approach cannot utilize these labels. The ABC is a
probabilistic binary classifier that effectively exploits the label
information, where normal data points are modeled using the AE as a component.
By maximizing the likelihood, the AE in the proposed ABC is trained to minimize
the reconstruction error for normal data points, and to maximize it for known
anomalies. Since our approach becomes able to reconstruct the normal data
points accurately and fails to reconstruct the known and unknown anomalies, it
can accurately discriminate both known and unknown anomalies from normal data
points. Experimental results show that the ABC achieves higher detection
performance than existing supervised and unsupervised methods.
| stat |
Telescoping Density-Ratio Estimation | Density-ratio estimation via classification is a cornerstone of unsupervised
learning. It has provided the foundation for state-of-the-art methods in
representation learning and generative modelling, with the number of use-cases
continuing to proliferate. However, it suffers from a critical limitation: it
fails to accurately estimate ratios p/q for which the two densities differ
significantly. Empirically, we find this occurs whenever the KL divergence
between p and q exceeds tens of nats. To resolve this limitation, we introduce
a new framework, telescoping density-ratio estimation (TRE), that enables the
estimation of ratios between highly dissimilar densities in high-dimensional
spaces. Our experiments demonstrate that TRE can yield substantial improvements
over existing single-ratio methods for mutual information estimation,
representation learning and energy-based modelling.
| stat |
Causal mediation analysis decomposition of between-hospital variance | Causal variance decompositions for a given disease-specific quality indicator
can be used to quantify differences in performance between hospitals or health
care providers. While variance decompositions can demonstrate variation in
quality of care, causal mediation analysis can be used to study care pathways
leading to the differences in performance between the institutions. This raises
the question of whether the two approaches can be combined to decompose
between-hospital variation in an outcome type indicator to that mediated
through a given process (indirect effect) and remaining variation due to all
other pathways (direct effect). For this purpose, we derive a causal mediation
analysis decomposition of between-hospital variance, discuss its
interpretation, and propose an estimation approach based on generalized linear
mixed models for the outcome and the mediator. We study the performance of the
estimators in a simulation study and demonstrate its use in administrative data
on kidney cancer care in Ontario.
| stat |
Mixture-based estimation of entropy | The entropy is a measure of uncertainty that plays a central role in
information theory. When the distribution of the data is unknown, an estimate
of the entropy needs be obtained from the data sample itself. We propose a
semi-parametric estimate, based on a mixture model approximation of the
distribution of interest. The estimate can rely on any type of mixture, but we
focus on Gaussian mixture model to demonstrate its accuracy and versatility.
Performance of the proposed approach is assessed through a series of simulation
studies. We also illustrate its use on two real-life data examples.
| stat |
Modeling continuous-time stochastic processes using $\mathcal{N}$-Curve
mixtures | Representations of sequential data are commonly based on the assumption that
observed sequences are realizations of an unknown underlying stochastic
process, where the learning problem includes determination of the model
parameters. In this context the model must be able to capture the multi-modal
nature of the data, without blurring between modes. This property is essential
for applications like trajectory prediction or human motion modeling. Towards
this end, a neural network model for continuous-time stochastic processes
usable for sequence prediction is proposed. The model is based on Mixture
Density Networks using B\'ezier curves with Gaussian random variables as
control points (abbrev.: $\mathcal{N}$-Curves). Key advantages of the model
include the ability of generating smooth multi-mode predictions in a single
inference step which reduces the need for Monte Carlo simulation, as required
in many multi-step prediction models, based on state-of-the-art neural
networks. Essential properties of the proposed approach are illustrated by
several toy examples and the task of multi-step sequence prediction. Further,
the model performance is evaluated on two real world use-cases, i.e. human
trajectory prediction and human motion modeling, outperforming different
state-of-the-art models.
| stat |
A bimodal gamma distribution: Properties, regression model and
applications | In this paper we propose a bimodal gamma distribution using a quadratic
transformation based on the alpha-skew-normal model. We discuss several
properties of this distribution such as mean, variance, moments, hazard rate
and entropy measures. Further, we propose a new regression model with censored
data based on the bimodal gamma distribution. This regression model can be very
useful to the analysis of real data and could give more realistic fits than
other special regression models. Monte Carlo simulations were performed to
check the bias in the maximum likelihood estimation. The proposed models are
applied to two real data sets found in literature.
| stat |
Deep Mixtures of Unigrams for uncovering Topics in Textual Data | Mixtures of Unigrams are one of the simplest and most efficient tools for
clustering textual data, as they assume that documents related to the same
topic have similar distributions of terms, naturally described by Multinomials.
When the classification task is particularly challenging, such as when the
document-term matrix is high-dimensional and extremely sparse, a more composite
representation can provide better insight on the grouping structure. In this
work, we developed a deep version of mixtures of Unigrams for the unsupervised
classification of very short documents with a large number of terms, by
allowing for models with further deeper latent layers; the proposal is derived
in a Bayesian framework. The behaviour of the Deep Mixtures of Unigrams is
empirically compared with that of other traditional and state-of-the-art
methods, namely $k$-means with cosine distance, $k$-means with Euclidean
distance on data transformed according to Semantic Analysis, Partition Around
Medoids, Mixture of Gaussians on semantic-based transformed data, hierarchical
clustering according to Ward's method with cosine dissimilarity, Latent
Dirichlet Allocation, Mixtures of Unigrams estimated via the EM algorithm,
Spectral Clustering and Affinity Propagation clustering. The performance is
evaluated in terms of both correct classification rate and Adjusted Rand Index.
Simulation studies and real data analysis prove that going deep in clustering
such data highly improves the classification accuracy.
| stat |
Strictly Batch Imitation Learning by Energy-based Distribution Matching | Consider learning a policy purely on the basis of demonstrated behavior --
that is, with no access to reinforcement signals, no knowledge of transition
dynamics, and no further interaction with the environment. This *strictly batch
imitation learning* problem arises wherever live experimentation is costly,
such as in healthcare. One solution is simply to retrofit existing algorithms
for apprenticeship learning to work in the offline setting. But such an
approach leans heavily on off-policy evaluation or offline model estimation,
and can be indirect and inefficient. We argue that a good solution should be
able to explicitly parameterize a policy (i.e. respecting action conditionals),
implicitly learn from rollout dynamics (i.e. leveraging state marginals), and
-- crucially -- operate in an entirely offline fashion. To address this
challenge, we propose a novel technique by *energy-based distribution matching*
(EDM): By identifying parameterizations of the (discriminative) model of a
policy with the (generative) energy function for state distributions, EDM
yields a simple but effective solution that equivalently minimizes a divergence
between the occupancy measure for the demonstrator and a model thereof for the
imitator. Through experiments with application to control and healthcare
settings, we illustrate consistent performance gains over existing algorithms
for strictly batch imitation learning.
| stat |
Function-on-Function Regression for the Identification of Epigenetic
Regions Exhibiting Windows of Susceptibility to Environmental Exposures | The ability to identify time periods when individuals are most susceptible to
exposures, as well as the biological mechanisms through which these exposures
act, is of great public health interest. Growing evidence supports an
association between prenatal exposure to air pollution and epigenetic marks,
such as DNA methylation, but the timing and gene-specific effects of these
epigenetic changes are not well understood. Here, we present the first study
that aims to identify prenatal windows of susceptibility to air pollution
exposures in cord blood DNA methylation. In particular, we propose a
function-on-function regression model that leverages data from nearby DNA
methylation probes to identify epigenetic regions that exhibit windows of
susceptibility to ambient particulate matter less than 2.5 microns
(PM$_{2.5}$). By incorporating the covariance structure among both the
multivariate DNA methylation outcome and the time-varying exposure under study,
this framework yields greater power to detect windows of susceptibility and
greater control of false discoveries than methods that model probes
independently. We compare our method to a distributed lag model approach that
models DNA methylation in a probe-by-probe manner, both in simulation and by
application to motivating data from the Project Viva birth cohort. In two
epigenetic regions selected based on prior studies of air pollution effects on
epigenome-wide methylation, we identify windows of susceptibility to PM$_{2.5}$
exposure near the beginning and middle of the third trimester of pregnancy.
| stat |
Stochastic Gradient Hamiltonian Monte Carlo for Non-Convex Learning | Stochastic Gradient Hamiltonian Monte Carlo (SGHMC) is a momentum version of
stochastic gradient descent with properly injected Gaussian noise to find a
global minimum. In this paper, non-asymptotic convergence analysis of SGHMC is
given in the context of non-convex optimization, where subsampling techniques
are used over an i.i.d dataset for gradient updates. Our results complement
those of [RRT17] and improve on those of [GGZ18].
| stat |
Parameter estimation in nonlinear mixed effect models based on ordinary
differential equations: an optimal control approach | We present a parameter estimation method for nonlinear mixed effect models
based on ordinary differential equations (NLME-ODEs). The method presented here
aims at regularizing the estimation problem in presence of model
misspecifications, practical identifiability issues and unknown initial
conditions. For doing so, we define our estimator as the minimizer of a cost
function which incorporates a possible gap between the assumed model at the
population level and the specific individual dynamic. The cost function
computation leads to formulate and solve optimal control problems at the
subject level. This control theory approach allows to bypass the need to know
or estimate initial conditions for each subject and it regularizes the
estimation problem in presence of poorly identifiable parameters. Comparing to
maximum likelihood, we show on simulation examples that our method improves
estimation accuracy in possibly partially observed systems with unknown initial
conditions or poorly identifiable parameters with or without model error. We
conclude this work with a real application on antibody concentration data after
vaccination against Ebola virus coming from phase 1 trials. We use the
estimated model discrepancy at the subject level to analyze the presence of
model misspecification.
| stat |
Spatial+: a novel approach to spatial confounding | In spatial regression models, collinearity between covariates and spatial
effects can lead to significant bias in effect estimates. This problem, known
as spatial confounding, is encountered modelling forestry data to assess the
effect of temperature on tree health. Reliable inference is difficult as
results depend on whether or not spatial effects are included in the model. The
mechanism behind spatial confounding is poorly understood and methods for
dealing with it are limited. We propose a novel approach, spatial+, in which
collinearity is reduced by replacing the covariates in the spatial model by
their residuals after spatial dependence has been regressed away. Using a thin
plate spline model formulation, we recognise spatial confounding as a
smoothing-induced bias identified by Rice (1986), and through asymptotic
analysis of the effect estimates, we show that spatial+ avoids the bias
problems of the spatial model. This is also demonstrated in a simulation study.
Spatial+ is straight-forward to implement using existing software and, as the
response variable is the same as that of the spatial model, standard model
selection criteria can be used for comparisons. A major advantage of the method
is also that it extends to models with non-Gaussian response distributions.
Finally, while our results are derived in a thin plate spline setting, the
spatial+ methodology transfers easily to other spatial model formulations.
| stat |
Projection pursuit based on Gaussian mixtures and evolutionary
algorithms | We propose a projection pursuit (PP) algorithm based on Gaussian mixture
models (GMMs). The negentropy obtained from a multivariate density estimated by
GMMs is adopted as the PP index to be maximised. For a fixed dimension of the
projection subspace, the GMM-based density estimation is projected onto that
subspace, where an approximation of the negentropy for Gaussian mixtures is
computed. Then, Genetic Algorithms (GAs) are used to find the optimal,
orthogonal projection basis by maximising the former approximation. We show
that this semi-parametric approach to PP is flexible and allows highly
informative structures to be detected, by projecting multivariate datasets onto
a subspace, where the data can be feasibly visualised. The performance of the
proposed approach is shown on both artificial and real datasets.
| stat |
False Discovery Rates to Detect Signals from Incomplete Spatially
Aggregated Data | There are a number of ways to test for the absence/presence of a spatial
signal in a completely observed fine-resolution image. One of these is a
powerful nonparametric procedure called Enhanced False Discovery Rate (EFDR). A
drawback of EFDR is that it requires the data to be defined on regular pixels
in a rectangular spatial domain. Here, we develop an EFDR procedure for
possibly incomplete data defined on irregular small areas. Motivated by
statistical learning, we use conditional simulation (CS) to condition on the
available data and simulate the full rectangular image at its finest resolution
many times (M, say). EFDR is then applied to each of these simulations
resulting in M estimates of the signal and M statistically dependent p-values.
Averaging over these estimates yields a single, combined estimate of a possible
signal, but inference is needed to determine whether there really is a signal
present. We test the original null hypothesis of no signal by combining the M
p-values into a single p-value using copulas and a composite likelihood. If the
null hypothesis of no signal is rejected, we use the combined estimate. We call
this new procedure EFDR-CS and, to demonstrate its effectiveness, we show
results from a simulation study; an experiment where we introduce aggregation
and incompleteness into temperature-change data in the Asia-Pacific; and an
application to total-column carbon dioxide from satellite remote sensing data
over a region of the Middle East, Afghanistan, and the western part of
Pakistan.
| stat |
Null-free False Discovery Rate Control Using Decoy Permutations | The traditional approaches to false discovery rate (FDR) control in multiple
hypothesis testing are usually based on the null distribution of a test
statistic. However, all types of null distributions, including the theoretical,
permutation-based and empirical ones, have some inherent drawbacks. For
example, the theoretical null might fail because of improper assumptions on the
sample distribution. Here, we propose a null distribution-free approach to FDR
control for multiple hypothesis testing. This approach, named target-decoy
procedure, simply builds on the ordering of tests by some statistic or score,
the null distribution of which is not required to be known. Competitive decoy
tests are constructed from permutations of original samples and are used to
estimate the false target discoveries. We prove that this approach controls the
FDR when the statistics are independent between different tests. Simulation
demonstrates that it is more stable and powerful than two existing popular
approaches. Evaluation is also made on a real dataset.
| stat |
Gaussian graphical models with graph constraints for magnetic moment
interaction in high entropy alloys | This article is motivated by studying the interaction of magnetic moments in
high entropy alloys (HEAs), which plays an important role in guiding HEA
designs in materials science. While first principles simulations can capture
magnetic moments of individual atoms, explicit models are required to analyze
their interactions. This is essentially an inverse covariance matrix estimation
problem. Unlike most of the literature on graphical models, the inverse
covariance matrix possesses inherent structural constraints encompassing node
types, topological distance of nodes, and partially specified conditional
dependence patterns. The present article is, to our knowledge, the first to
consider such intricate structures in graphical models. In particular, we
introduce graph constraints to formulate these structures that arise from
domain knowledge and are critical for interpretability, which leads to a
Bayesian conditional autoregressive model with graph constraints (CARGO) for
structured inverse covariance matrices. The CARGO method enjoys efficient
implementation with a modified Gauss-Seidel scheme through proximity operators
for closed-form posterior exploration. We establish algorithmic convergence for
the proposed algorithm under a verifiable stopping criterion. Simulations show
competitive performance of CARGO relative to several other methods and confirm
our convergence analysis. In a novel real data application to HEAs, the
proposed methods lead to data-driven quantification and interpretation of
magnetic moment interactions with high tractability and transferability.
| stat |
Harmonised LUCAS in-situ data and photos on land cover and use from 5
tri-annual surveys in the European Union | Accurately characterizing land surface changes with Earth Observation
requires geo-localized ground truth. In the European Union (EU), a tri-annual
surveyed sample of land cover and land use has been collected since 2006 under
the Land Use/Cover Area frame Survey (LUCAS). A total of 1,351,293 observations
at 651,780 unique locations for 117 variables along with 5.4 million photos
were collected during five LUCAS surveys. Until now, these data have never been
harmonised into one database, limiting full exploitation of the information.
This paper describes the LUCAS point sampling/surveying methodology, including
collection of standard variables such as land cover, environmental parameters,
and full resolution landscape and point photos, and then describes the
harmonisation process. The resulting harmonised database is the most
comprehensive in-situ dataset on land cover and use in the EU. The database is
valuable for geo-spatial and statistical analysis of land use and land cover
change. Furthermore, its potential to provide multi-temporal in-situ data will
be enhanced by recent computational advances such as deep learning.
| stat |
Systematic errors in estimates of $R_t$ from symptomatic cases in the
presence of observation bias | We consider the problem of estimating the reproduction number $R_t$ of an
epidemic for populations where the probability of detection of cases depends on
a known covariate. We argue that in such cases the normal empirical estimator
can fail when the prevalence of cases among groups changes with time. We
propose a Bayesian strategy to resolve the problem, as well as a simple
solution in the case of large number of cases. We illustrate the issue and its
solution on a simple yet realistic simulation study, and discuss the general
relevance of the issue to the current covid19 pandemic.
| stat |
Using Invalid Instruments on Purpose: Focused Moment Selection and
Averaging for GMM | In finite samples, the use of a slightly endogenous but highly relevant
instrument can reduce mean-squared error (MSE). Building on this observation, I
propose a novel moment selection procedure for GMM -- the Focused Moment
Selection Criterion (FMSC) -- in which moment conditions are chosen not based
on their validity but on the MSE of their associated estimator of a
user-specified target parameter. The FMSC mimics the situation faced by an
applied researcher who begins with a set of relatively mild "baseline"
assumptions and must decide whether to impose any of a collection of stronger
but more controversial "suspect" assumptions. When the (correctly specified)
baseline moment conditions identify the model, the FMSC provides an
asymptotically unbiased estimator of asymptotic MSE, allowing us to select over
the suspect moment conditions. I go on to show how the framework used to derive
the FMSC can address the problem of inference post-moment selection. Treating
post-selection estimators as a special case of moment-averaging, in which
estimators based on different moment sets are given data-dependent weights, I
propose simulation-based procedures for inference that can be applied to a
variety of formal and informal moment-selection and averaging procedures. Both
the FMSC and confidence interval procedures perform well in simulations. I
conclude with an empirical example examining the effect of instrument selection
on the estimated relationship between malaria and income per capita.
| stat |
Covariance Prediction via Convex Optimization | We consider the problem of predicting the covariance of a zero mean Gaussian
vector, based on another feature vector. We describe a covariance predictor
that has the form of a generalized linear model, i.e., an affine function of
the features followed by an inverse link function that maps vectors to
symmetric positive definite matrices. The log-likelihood is a concave function
of the predictor parameters, so fitting the predictor involves convex
optimization. Such predictors can be combined with others, or recursively
applied to improve performance.
| stat |
Empirical Decision Rules for Improving the Uncertainty Reporting of
Small Sample System Usability Scale Scores | The System Usability Scale (SUS) is a short, survey-based approach used to
determine the usability of a system from an end user perspective once a
prototype is available for assessment. Individual scores are gathered using a
10-question survey with the survey results reported in terms of central
tendency (sample mean) as an estimate of the system's usability (the SUS study
score), and confidence intervals on the sample mean are used to communicate
uncertainty levels associated with this point estimate. When the number of
individuals surveyed is large, the SUS study scores and accompanying confidence
intervals relying upon the central limit theorem for support are appropriate.
However, when only a small number of users are surveyed, reliance on the
central limit theorem falls short, resulting in confidence intervals that
suffer from parameter bound violations and interval widths that confound
mappings to adjective and other constructed scales. These shortcomings are
especially pronounced when the underlying SUS score data is skewed, as it is in
many instances. This paper introduces an empirically-based remedy for such
small-sample circumstances, proposing a set of decision rules that leverage
either an extended bias-corrected accelerated (BCa) bootstrap confidence
interval or an empirical Bayesian credibility interval about the sample mean to
restore and bolster subsequent confidence interval accuracy. Data from
historical SUS assessments are used to highlight shortfalls in current
practices and to demonstrate the improvements these alternate approaches offer
while remaining statistically defensible. A freely available, online
application is introduced and discussed that automates SUS analysis under these
decision rules, thereby assisting usability practitioners in adopting the
advocated approaches.
| stat |
Autoregressive Mixture Models for Serial Correlation Clustering of Time
Series Data | Clustering individuals into similar groups in longitudinal studies can
improve time series models by combining information across like individuals.
While there is a well developed literature for clustering of time series, these
approaches tend to generate clusters independently of the model training
procedure which can lead to poor model fit. We propose a novel method that
simultaneously clusters and fits autoregression models for groups of similar
individuals. We apply a Wishart mixture model so as to cluster individuals
while modeling the corresponding autocorrelation matrices at the same time. The
fitted Wishart scale matrices map to cluster-level autoregressive coefficients
through the Yule-Walker equations, fitting robust parsimonious autoregressive
mixture models. This approach is able to discern differences in underlying
serial variation of time series while accounting for an individual's intrinsic
variability. We prove consistency of our cluster membership estimator and
compare our approach against competing methods through simulation as well as by
modeling regional COVID-19 infection rates.
| stat |
BERT based patent novelty search by training claims to their own
description | In this paper we present a method to concatenate patent claims to their own
description. By applying this method, BERT trains suitable descriptions for
claims. Such a trained BERT (claim-to-description- BERT) could be able to
identify novelty relevant descriptions for patents. In addition, we introduce a
new scoring scheme, relevance scoring or novelty scoring, to process the output
of BERT in a meaningful way. We tested the method on patent applications by
training BERT on the first claims of patents and corresponding descriptions.
BERT's output has been processed according to the relevance score and the
results compared with the cited X documents in the search reports. The test
showed that BERT has scored some of the cited X documents as highly relevant.
| stat |
Isotonic Distributional Regression | Isotonic distributional regression (IDR) is a powerful nonparametric
technique for the estimation of conditional distributions under order
restrictions. In a nutshell, IDR learns conditional distributions that are
calibrated, and simultaneously optimal relative to comprehensive classes of
relevant loss functions, subject to isotonicity constraints in terms of a
partial order on the covariate space. Nonparametric isotonic quantile
regression and nonparametric isotonic binary regression emerge as special
cases. For prediction, we propose an interpolation method that generalizes
extant specifications under the pool adjacent violators algorithm. We recommend
the use of IDR as a generic benchmark technique in probabilistic forecast
problems, as it does not involve any parameter tuning nor implementation
choices, except for the selection of a partial order on the covariate space.
The method can be combined with subsample aggregation, with the benefits of
smoother regression functions and gains in computational efficiency. In a
simulation study, we compare methods for distributional regression in terms of
the continuous ranked probability score (CRPS) and $L_2$ estimation error,
which are closely linked. In a case study on raw and postprocessed quantitative
precipitation forecasts from a leading numerical weather prediction system, IDR
is competitive with state of the art techniques.
| stat |
Modeling of Electrical Resistivity of Soil Based on Geotechnical
Properties | Determining the relationship between the electrical resistivity of soil and
its geotechnical properties is an important engineering problem. This study
aims to develop methodology for finding the best model that can be used to
predict the electrical resistivity of soil, based on knowing its geotechnical
properties. The research develops several linear models, three non-linear
models, and three artificial neural network models (ANN). These models are
applied to the experimental data set comprises 864 observations and five
variables. The results show that there are significant exponential negative
relationships between the electrical resistivity of soil and its geotechnical
properties. The most accurate prediction values are obtained using the ANN
model. The cross-validation analysis confirms the high precision of the
selected predictive model. This research is the first rigorous systematic
analysis and comparison of difference methodologies in ground electrical
resistivity studies. It provides practical guidelines and examples of design,
development and testing non-linear relationships in engineering intelligent
systems and applications.
| stat |
AKM$^2$D : An Adaptive Framework for Online Sensing and Anomaly
Quantification | In point-based sensing systems such as coordinate measuring machines (CMM)
and laser ultrasonics where complete sensing is impractical due to the high
sensing time and cost, adaptive sensing through a systematic exploration is
vital for online inspection and anomaly quantification. Most of the existing
sequential sampling methodologies focus on reducing the overall fitting error
for the entire sampling space. However, in many anomaly quantification
applications, the main goal is to estimate sparse anomalous regions in the
pixel-level accurately. In this paper, we develop a novel framework named
Adaptive Kernelized Maximum-Minimum Distance AKM$^2$D to speed up the
inspection and anomaly detection process through an intelligent sequential
sampling scheme integrated with fast estimation and detection. The proposed
method balances the sampling efforts between the space-filling sampling
(exploration) and focused sampling near the anomalous region (exploitation).
The proposed methodology is validated by conducting simulations and a case
study of anomaly detection in composite sheets using a guided wave test.
| stat |
Why ResNet Works? Residuals Generalize | Residual connections significantly boost the performance of deep neural
networks. However, there are few theoretical results that address the influence
of residuals on the hypothesis complexity and the generalization ability of
deep neural networks. This paper studies the influence of residual connections
on the hypothesis complexity of the neural network in terms of the covering
number of its hypothesis space. We prove that the upper bound of the covering
number is the same as chain-like neural networks, if the total numbers of the
weight matrices and nonlinearities are fixed, no matter whether they are in the
residuals or not. This result demonstrates that residual connections may not
increase the hypothesis complexity of the neural network compared with the
chain-like counterpart. Based on the upper bound of the covering number, we
then obtain an $\mathcal O(1 / \sqrt{N})$ margin-based multi-class
generalization bound for ResNet, as an exemplary case of any deep neural
network with residual connections. Generalization guarantees for similar
state-of-the-art neural network architectures, such as DenseNet and ResNeXt,
are straight-forward. From our generalization bound, a practical implementation
is summarized: to approach a good generalization ability, we need to use
regularization terms to control the magnitude of the norms of weight matrices
not to increase too much, which justifies the standard technique of weight
decay.
| stat |
First digit law from Laplace transform | The occurrence of digits 1 through 9 as the leftmost nonzero digit of numbers
from real-world sources is distributed unevenly according to an empirical law,
known as Benford's law or the first digit law. It remains obscure why a variety
of data sets generated from quite different dynamics obey this particular law.
We perform a study of Benford's law from the application of the Laplace
transform, and find that the logarithmic Laplace spectrum of the digital
indicator function can be approximately taken as a constant. This particular
constant, being exactly the Benford term, explains the prevalence of Benford's
law. The slight variation from the Benford term leads to deviations from
Benford's law for distributions which oscillate violently in the inverse
Laplace space. We prove that the whole family of completely monotonic
distributions can satisfy Benford's law within a small bound. Our study
suggests that Benford's law originates from the way that we write numbers, thus
should be taken as a basic mathematical knowledge.
| stat |
Goodness-of-fit testing in high-dimensional generalized linear models | We propose a family of tests to assess the goodness-of-fit of a
high-dimensional generalized linear model. Our framework is flexible and may be
used to construct an omnibus test or directed against testing specific
non-linearities and interaction effects, or for testing the significance of
groups of variables. The methodology is based on extracting left-over signal in
the residuals from an initial fit of a generalized linear model. This can be
achieved by predicting this signal from the residuals using modern flexible
regression or machine learning methods such as random forests or boosted trees.
Under the null hypothesis that the generalized linear model is correct, no
signal is left in the residuals and our test statistic has a Gaussian limiting
distribution, translating to asymptotic control of type I error. Under a local
alternative, we establish a guarantee on the power of the test. We illustrate
the effectiveness of the methodology on simulated and real data examples by
testing goodness-of-fit in logistic regression models. Software implementing
the methodology is available in the R package `GRPtests'.
| stat |
Causal inference with confounders missing not at random | It is important to draw causal inference from observational studies, which,
however, becomes challenging if the confounders have missing values. Generally,
causal effects are not identifiable if the confounders are missing not at
random. We propose a novel framework to nonparametrically identify causal
effects with confounders subject to an outcome-independent missingness, that
is, the missing data mechanism is independent of the outcome, given the
treatment and possibly missing confounders. We then propose a nonparametric
two-stage least squares estimator and a parametric estimator for causal
effects.
| stat |
Multilinear Common Component Analysis via Kronecker Product
Representation | We consider the problem of extracting a common structure from multiple tensor
datasets. For this purpose, we propose multilinear common component analysis
(MCCA) based on Kronecker products of mode-wise covariance matrices. MCCA
constructs a common basis represented by linear combinations of the original
variables which loses as little information of the multiple tensor datasets. We
also develop an estimation algorithm for MCCA that guarantees mode-wise global
convergence. Numerical studies are conducted to show the effectiveness of MCCA.
| stat |
Bayesian Algorithms for Decentralized Stochastic Bandits | We study a decentralized cooperative multi-agent multi-armed bandit problem
with $K$ arms and $N$ agents connected over a network. In our model, each arm's
reward distribution is same for all agents, and rewards are drawn independently
across agents and over time steps. In each round, agents choose an arm to play
and subsequently send a message to their neighbors. The goal is to minimize
cumulative regret averaged over the entire network. We propose a decentralized
Bayesian multi-armed bandit framework that extends single-agent Bayesian bandit
algorithms to the decentralized setting. Specifically, we study an information
assimilation algorithm that can be combined with existing Bayesian algorithms,
and using this, we propose a decentralized Thompson Sampling algorithm and
decentralized Bayes-UCB algorithm. We analyze the decentralized Thompson
Sampling algorithm under Bernoulli rewards and establish a problem-dependent
upper bound on the cumulative regret. We show that regret incurred scales
logarithmically over the time horizon with constants that match those of an
optimal centralized agent with access to all observations across the network.
Our analysis also characterizes the cumulative regret in terms of the network
structure. Through extensive numerical studies, we show that our extensions of
Thompson Sampling and Bayes-UCB incur lesser cumulative regret than the
state-of-art algorithms inspired by the Upper Confidence Bound algorithm. We
implement our proposed decentralized Thompson Sampling under gossip protocol,
and over time-varying networks, where each communication link has a fixed
probability of failure.
| stat |
EventDetectR -- An Open-Source Event Detection System | EventDetectR: An efficient Event Detection System (EDS) capable of detecting
unexpected water quality conditions. This approach uses multiple algorithms to
model the relationship between various multivariate water quality signals. Then
the residuals of the models were utilized in constructing the event detection
algorithm, which provides a continuous measure of the probability of an event
at every time step. The proposed framework was tested for water contamination
events with industrial data from automated water quality sensors. The results
showed that the framework is reliable with better performance and is highly
suitable for event detection.
| stat |
Fair tournament design: A flaw of the UEFA Euro 2020 qualification | The integrity of a sport can be seriously undermined if its rules punish
winning as this creates incentives for strategic manipulation. Therefore, a
sports tournament can be called unfair if the overall win probabilities are not
ordered according to the teams' ranking based on their past performances. We
present how statistical methods can contribute to choosing a tournament format
that is in line with the above axiom. In particular, the qualification for the
2020 UEFA European Championship is shown to violate this requirement: being a
top team in the lowest-ranked League D of the 2018/19 UEFA Nations League
substantially increases the probability of qualifying compared to being a
bottom team in the higher-ranked League C. The unfairness can be remarkably
reduced or even eliminated with slightly changing the path formation policy of
the UEFA Euro 2020 qualifying play-offs. The misaligned design has severely
punished a team for winning a match years before. Since the deficiency is an
inherent feature of the qualifying process, the Union of European Football
Associations (UEFA) should reconsider the format of future tournaments to
eliminate the unfair advantage enjoyed by certain teams.
| stat |
R-optimal designs for multi-response regression models with
multi-factors | We investigate R-optimal designs for multi-response regression models with
multi-factors, where the random errors in these models are correlated. Several
theoretical results are derived for Roptimal designs, including scale
invariance, reflection symmetry, line and plane symmetry, and dependence on the
covariance matrix of the errors. All the results can be applied to linear and
nonlinear models. In addition, an efficient algorithm based on an interior
point method is developed for finding R-optimal designs on discrete design
spaces. The algorithm is very flexible, and can be applied to any
multi-response regression model.
| stat |
New multivariate tests for assessing covariate balance in matched
observational studies | We propose new tests for assessing whether covariates in a treatment group
and matched control group are balanced in observational studies. The tests
exhibit high power under a wide range of multivariate alternatives, some of
which existing tests have little power for. The asymptotic permutation null
distributions of the proposed tests are studied and the p-values calculated
through the asymptotic results work well in finite samples, facilitating the
application of the test to large data sets. The tests are illustrated in a
study of the effect of smoking on blood lead levels. The proposed tests are
implemented in an R package BalanceCheck.
| stat |
Computing Accurate Probabilistic Estimates of One-D Entropy from
Equiprobable Random Samples | We develop a simple Quantile Spacing (QS) method for accurate probabilistic
estimation of one-dimensional entropy from equiprobable random samples, and
compare it with the popular Bin-Counting (BC) method. In contrast to BC, which
uses equal-width bins with varying probability mass, the QS method uses
estimates of the quantiles that divide the support of the data generating
probability density function (pdf) into equal-probability-mass intervals.
Whereas BC requires optimal tuning of a bin-width hyper-parameter whose value
varies with sample size and shape of the pdf, QS requires specification of the
number of quantiles to be used. Results indicate, for the class of
distributions tested, that the optimal number of quantile-spacings is a fixed
fraction of the sample size (empirically determined to be ~0.25-0.35), and that
this value is relatively insensitive to distributional form or sample size,
providing a clear advantage over BC since hyperparameter tuning is not
required. Bootstrapping is used to approximate the sampling variability
distribution of the resulting entropy estimate, and is shown to accurately
reflect the true uncertainty. For the four distributional forms studied
(Gaussian, Log-Normal, Exponential and Bimodal Gaussian Mixture), expected
estimation bias is less than 1% and uncertainty is relatively low even for very
small sample sizes. We speculate that estimating quantile locations, rather
than bin-probabilities, results in more efficient use of the information in the
data to approximate the underlying shape of an unknown data generating pdf.
| stat |
Distinction of groups of gamma-ray bursts in the BATSE catalog through
fuzzy clustering | In search for the possible astrophysical sources behind origination of the
diverse gamma-ray bursts, cluster analyses are performed to find homogeneous
groups, which discover an intermediate group other than the conventional short
and long bursts. However, very recently, few studies indicate a possibility of
the existence of more than three (namely five) groups. Therefore, in this
paper, fuzzy clustering is conducted on the gamma-ray bursts from the final
'Burst and Transient Source Experiment' catalog to cross-check the significance
of these new groups. Meticulous study on individual bursts based on their
memberships in the fuzzy clusters confirms the previously well-known three
groups against the newly found five.
| stat |
Applying the Network Item Response Model to Student Assessment Data | This study discusses an alternative tool for modeling student assessment
data. The model constructs networks from a matrix item responses and attempts
to represent these data in low dimensional Euclidean space. This procedure has
advantages over common methods used for modeling student assessment data such
as Item Response Theory because it relaxes the highly restrictive
local-independence assumption. This article provides a deep discussion of the
model and the steps one must take to estimate it. To enable extending a present
model by adding data, two methods for estimating the positions of new
individuals in the network are discussed. Then, a real data analysis is then
provided as a case study on using the model and how to interpret the results.
Finally, the model is compared and contrasted to other popular models in
psychological and educational measurement: Item response theory (IRT) and
network psychometric Ising model for binary data.
| stat |
The covariance matrix of Green's functions and its application to
machine learning | In this paper, a regression algorithm based on Green's function theory is
proposed and implemented. We first survey Green's function for the Dirichlet
boundary value problem of 2nd order linear ordinary differential equation,
which is a reproducing kernel of a suitable Hilbert space. We next consider a
covariance matrix composed of the normalized Green's function, which is
regarded as aprobability density function. By supporting Bayesian approach, the
covariance matrix gives predictive distribution, which has the predictive mean
$\mu$ and the confidence interval [$\mu$-2s, $\mu$+2s], where s stands for a
standard deviation.
| stat |
Identification of Matrix Joint Block Diagonalization | Given a set $\mathcal{C}=\{C_i\}_{i=1}^m$ of square matrices, the matrix
blind joint block diagonalization problem (BJBDP) is to find a full column rank
matrix $A$ such that $C_i=A\Sigma_iA^\text{T}$ for all $i$, where $\Sigma_i$'s
are all block diagonal matrices with as many diagonal blocks as possible. The
BJBDP plays an important role in independent subspace analysis (ISA). This
paper considers the identification problem for BJBDP, that is, under what
conditions and by what means, we can identify the diagonalizer $A$ and the
block diagonal structure of $\Sigma_i$, especially when there is noise in
$C_i$'s. In this paper, we propose a ``bi-block diagonalization'' method to
solve BJBDP, and establish sufficient conditions under which the method is able
to accomplish the task. Numerical simulations validate our theoretical results.
To the best of the authors' knowledge, existing numerical methods for BJBDP
have no theoretical guarantees for the identification of the exact solution,
whereas our method does.
| stat |
Geometric goodness of fit measure to detect patterns in data point
clouds | We derived a geometric goodness-of-fit index, similar to $R^2$ using
topological data analysis techniques. We build the Vietoris-Rips complex from
the data-cloud projected onto each variable. Estimating the area of the complex
and their domain, we create an index that measures the emptiness of the space
with respect to the data. We made the analysis with an own package called TopSA
(Topological Sensitivy Analysis).
| stat |
Score-Driven Exponential Random Graphs: A New Class of Time-Varying
Parameter Models for Dynamical Networks | Motivated by the evidence that real-world networks evolve in time and may
exhibit non-stationary features, we propose an extension of the Exponential
Random Graph Models (ERGMs) accommodating the time variation of network
parameters. Within the ERGM framework, a network realization is sampled from a
static probability distribution defined parametrically in terms of network
statistics. Inspired by the fast growing literature on Dynamic Conditional
Score-driven models, in our approach, each parameter evolves according to an
updating rule driven by the score of the conditional distribution. We
demonstrate the flexibility of the score-driven ERGMs, both as data generating
processes and as filters, and we prove the advantages of the dynamic version
with respect to the static one. Our method captures dynamical network
dependencies, that emerge from the data, and allows for a test discriminating
between static or time-varying parameters. Finally, we corroborate our findings
with the application to networks from real financial and political systems
exhibiting non stationary dynamics.
| stat |
Online Estimation of Multiple Dynamic Graphs in Pattern Sequences | Sequences of correlated binary patterns can represent many time-series data
including text, movies, and biological signals. These patterns may be described
by weighted combinations of a few dominant structures that underpin specific
interactions among the binary elements. To extract the dominant correlation
structures and their contributions to generating data in a time-dependent
manner, we model the dynamics of binary patterns using the state-space model of
an Ising-type network that is composed of multiple undirected graphs. We
provide a sequential Bayes algorithm to estimate the dynamics of weights on the
graphs while gaining the graph structures online. This model can uncover
overlapping graphs underlying the data better than a traditional orthogonal
decomposition method, and outperforms an original time-dependent Ising model.
We assess the performance of the method by simulated data, and demonstrate that
spontaneous activity of cultured hippocampal neurons is represented by dynamics
of multiple graphs.
| stat |
Cross-validation based adaptive sampling for Gaussian process models | In many real-world applications, we are interested in approximating
black-box, costly functions as accurately as possible with the smallest number
of function evaluations. A complex computer code is an example of such a
function. In this work, a Gaussian process (GP) emulator is used to approximate
the output of complex computer code. We consider the problem of extending an
initial experiment (set of model runs) sequentially to improve the emulator.
A sequential sampling approach based on leave-one-out (LOO) cross-validation
is proposed that can be easily extended to a batch mode. This is a desirable
property since it saves the user time when parallel computing is available.
After fitting a GP to training data points, the expected squared LOO (ES-LOO)
error is calculated at each design point. ES-LOO is used as a measure to
identify important data points. More precisely, when this quantity is large at
a point it means that the quality of prediction depends a great deal on that
point and adding more samples nearby could improve the accuracy of the GP. As a
result, it is reasonable to select the next sample where ES-LOO is maximised.
However, ES-LOO is only known at the experimental design and needs to be
estimated at unobserved points. To do this, a second GP is fitted to the ES-LOO
errors and where the maximum of the modified expected improvement (EI)
criterion occurs is chosen as the next sample. EI is a popular acquisition
function in Bayesian optimisation and is used to trade-off between local/global
search. However, it has a tendency towards exploitation, meaning that its
maximum is close to the (current) "best" sample. To avoid clustering, a
modified version of EI, called pseudo expected improvement, is employed which
is more explorative than EI yet allows us to discover unexplored regions. Our
results show that the proposed sampling method is promising.
| stat |
A notion of depth for sparse functional data | Data depth is a well-known and useful nonparametric tool for analyzing
functional data. It provides a novel way of ranking a sample of curves from the
center outwards and defining robust statistics, such as the median or trimmed
means. It has also been used as a building block for functional outlier
detection methods and classification. Several notions of depth for functional
data were introduced in the literature in the last few decades. These
functional depths can only be directly applied to samples of curves measured on
a fine and common grid. In practice, this is not always the case, and curves
are often observed at sparse and subject dependent grids. In these scenarios
the usual approach consists in estimating the trajectories on a common dense
grid, and using the estimates in the depth analysis. This approach ignores the
uncertainty associated with the curves estimation step. Our goal is to extend
the notion of depth so that it takes into account this uncertainty. Using both
functional estimates and their associated confidence intervals, we propose a
new method that allows the curve estimation uncertainty to be incorporated into
the depth analysis. We describe the new approach using the modified band depth
although any other functional depth could be used. The performance of the
proposed methodology is illustrated using simulated curves in different
settings where we control the degree of sparsity. Also a real data set
consisting of female medflies egg-laying trajectories is considered. The
results show the benefits of using uncertainty when computing depth for sparse
functional data.
| stat |
Will Artificial Intelligence supersede Earth System and Climate Models? | We outline a perspective of an entirely new research branch in Earth and
climate sciences, where deep neural networks and Earth system models are
dismantled as individual methodological approaches and reassembled as learning,
self-validating, and interpretable Earth system model-network hybrids.
Following this path, we coin the term "Neural Earth System Modelling" (NESYM)
and highlight the necessity of a transdisciplinary discussion platform,
bringing together Earth and climate scientists, big data analysts, and AI
experts. We examine the concurrent potential and pitfalls of Neural Earth
System Modelling and discuss the open question whether artificial intelligence
will not only infuse Earth system modelling, but ultimately render them
obsolete.
| stat |
Beyond the EM Algorithm: Constrained Optimization Methods for Latent
Class Model | Latent class model (LCM), which is a finite mixture of different categorical
distributions, is one of the most widely used models in statistics and machine
learning fields. Because of its non-continuous nature and the flexibility in
shape, researchers in practice areas such as marketing and social sciences also
frequently use LCM to gain insights from their data. One likelihood-based
method, the Expectation-Maximization (EM) algorithm, is often used to obtain
the model estimators. However, the EM algorithm is well-known for its
notoriously slow convergence. In this research, we explore alternative
likelihood-based methods that can potential remedy the slow convergence of the
EM algorithm. More specifically, we regard likelihood-based approach as a
constrained nonlinear optimization problem, and apply quasi-Newton type methods
to solve them. We examine two different constrained optimization methods to
maximize the log likelihood function. We present simulation study results to
show that the proposed methods not only converge in less iterations than the EM
algorithm but also produce more accurate model estimators.
| stat |
L-moments for automatic threshold selection in extreme value analysis | In extreme value analysis, sensitivity of inference to the definition of
extreme event is a paramount issue. Under the peaks-over-threshold (POT)
approach, this translates directly into the need of fitting a Generalized
Pareto distribution to observations above a suitable level that balances bias
versus variance of estimates. Selection methodologies established in the
literature face recurrent challenges such as an inherent subjectivity or high
computational intensity. We suggest a truly automated method for threshold
detection, aiming at time efficiency and elimination of subjective judgment.
Based on the well-established theory of L-moments, this versatile data-driven
technique can handle batch processing of large collections of extremes data,
while also presenting good performance on small samples.
The technique's performance is evaluated in a large simulation study and
illustrated with significant wave height data sets from the literature. We find
that it compares favorably to other state-of-the-art methods regarding the
choice of threshold, associated parameter estimation and the ultimate goal of
computationally efficient return level estimation.
| stat |
Space-Time Landslide Predictive Modelling | Landslides are nearly ubiquitous phenomena and pose severe threats to people,
properties, and the environment. Investigators have for long attempted to
estimate landslide hazard to determine where, when, and how destructive
landslides are expected to be in an area. This information is useful to design
landslide mitigation strategies, and to reduce landslide risk and societal and
economic losses. In the geomorphology literature, most attempts at predicting
the occurrence of populations of landslides rely on the observation that
landslides are the result of multiple interacting, conditioning and triggering
factors. Here, we propose a novel Bayesian modelling framework for the
prediction of space-time landslide occurrences of the slide type caused by
weather triggers. We consider log-Gaussian cox processes, assuming that
individual landslides stem from a point process described by an unknown
intensity function. We tested our prediction framework in the Collazzone area,
Umbria, Central Italy, for which a detailed multi-temporal landslide inventory
spanning 1941-2014 is available together with lithological and bedding data. We
tested five models of increasing complexity. Our most complex model includes
fixed effects and latent spatio-temporal effects, thus largely fulfilling the
common definition of landslide hazard in the literature. We quantified the
spatio-temporal predictive skill of our model and found that it performed
better than simpler alternatives. We then developed a novel classification
strategy and prepared an intensity-susceptibility landslide map, providing more
information than traditional susceptibility zonations for land planning and
management. We expect our novel approach to lead to better projections of
future landslides, and to improve our collective understanding of the evolution
of landscapes dominated by mass-wasting processes under geophysical and weather
triggers.
| stat |
Scoring Predictions at Extreme Quantiles | Prediction of quantiles at extreme tails is of interest in numerous
applications. Extreme value modelling provides various competing predictors for
this point prediction problem. A common method of assessment of a set of
competing predictors is to evaluate their predictive performance in a given
situation. However, due to the extreme nature of this inference problem, it can
be possible that the predicted quantiles are not seen in the historical
records, particularly when the sample size is small. This situation poses a
problem to the validation of the prediction with its realisation. In this
article, we propose two non-parametric scoring approaches to assess extreme
quantile estimates. These methods are based on predicting a sequence of equally
extremal quantiles on different parts of the data. We then use the quantile
scoring function to evaluate the competing predictors. The performance of the
scoring methods is compared with the conventional scoring method and
superiority of the former methods are demonstrated in a simulation study. The
methods are then applied to reanalyse cyber Netflow data from Los Alamos
National Laboratory and daily precipitation data at a station in California
available from Global Historical Climatology Network.
| stat |
Bayesian Item Response Modeling in R with brms and Stan | Item Response Theory (IRT) is widely applied in the human sciences to model
persons' responses on a set of items measuring one or more latent constructs.
While several R packages have been developed that implement IRT models, they
tend to be restricted to respective prespecified classes of models. Further,
most implementations are frequentist while the availability of Bayesian methods
remains comparably limited. We demonstrate how to use the R package brms
together with the probabilistic programming language Stan to specify and fit a
wide range of Bayesian IRT models using flexible and intuitive multilevel
formula syntax. Further, item and person parameters can be related in both a
linear or non-linear manner. Various distributions for categorical, ordinal,
and continuous responses are supported. Users may even define their own custom
response distribution for use in the presented framework. Common IRT model
classes that can be specified natively in the presented framework include 1PL
and 2PL logistic models optionally also containing guessing parameters, graded
response and partial credit ordinal models, as well as drift diffusion models
of response times coupled with binary decisions. Posterior distributions of
item and person parameters can be conveniently extracted and post-processed.
Model fit can be evaluated and compared using Bayes factors and efficient
cross-validation procedures.
| stat |
Dose-response modeling in high-throughput cancer drug screenings: An
end-to-end approach | Personalized cancer treatments based on the molecular profile of a patient's
tumor are an emerging and exciting class of treatments in oncology. As genomic
tumor profiling is becoming more common, targeted treatments to specific
molecular alterations are gaining traction. To discover new potential
therapeutics that may apply to broad classes of tumors matching some molecular
pattern, experimentalists and pharmacologists rely on high-throughput, in-vitro
screens of many compounds against many different cell lines. We propose a
hierarchical Bayesian model of how cancer cell lines respond to drugs in these
experiments and develop a method for fitting the model to real-world
high-throughput screening data. Through a case study, the model is shown to
capture nontrivial associations between molecular features and drug response,
such as requiring both wild type TP53 and overexpression of MDM2 to be
sensitive to Nutlin-3(a). In quantitative benchmarks, the model outperforms a
standard approach in biology, with ~20% lower predictive error on held out
data. When combined with a conditional randomization testing procedure, the
model discovers biomarkers of therapeutic response that recapitulate known
biology and suggest new avenues for investigation. All code for the paper is
publicly available at https://github.com/tansey/deep-dose-response.
| stat |
Effectiveness and Compliance to Social Distancing During COVID-19 | In the absence of pharmaceutical interventions to curb the spread of
COVID-19, countries relied on a number of nonpharmaceutical interventions to
fight the first wave of the pandemic. The most prevalent one has been
stay-at-home orders, whose the goal is to limit the physical contact between
people, which consequently will reduce the number of secondary infections
generated. In this work, we use a detailed set of mobility data to evaluate the
impact that these interventions had on alleviating the spread of the virus in
the US as measured through the COVID-19-related deaths. To establish this
impact, we use the notion of Granger causality between two time-series. We show
that there is a unidirectional Granger causality, from the median percentage of
time spent daily at home to the daily number of COVID-19-related deaths with a
lag of 2 weeks. We further analyze the mobility patterns at the census block
level to identify which parts of the population might encounter difficulties in
adhering and complying with social distancing measures. This information is
important, since it can consequently drive interventions that aim at helping
these parts of the population.
| stat |
A note on breaking ties among sample medians | Given samples $x_1,\cdots,x_n$, it is well known that any sample median value
(not necessarily unique) minimizes the absolute loss $\sum_{i=1}^n |q-x_i|$.
Interestingly, we show that the minimizer of the loss
$\sum_{i=1}^n|q-x_i|^{1+\epsilon}$ exhibits a singular perturbation behaviour
that provides a unique definition for the sample median as $\epsilon
\rightarrow 0$. This definition is the unique point among all candidate median
values that balances the $logarithmic$ moment of the empirical distribution.
The result generalizes directly to breaking ties among sample quantiles when
the quantile regression loss is modified in the same way.
| stat |
Gaussian Mixture Clustering Using Relative Tests of Fit | We consider clustering based on significance tests for Gaussian Mixture
Models (GMMs). Our starting point is the SigClust method developed by Liu et
al. (2008), which introduces a test based on the k-means objective (with k = 2)
to decide whether the data should be split into two clusters. When applied
recursively, this test yields a method for hierarchical clustering that is
equipped with a significance guarantee. We study the limiting distribution and
power of this approach in some examples and show that there are large regions
of the parameter space where the power is low. We then introduce a new test
based on the idea of relative fit. Unlike prior work, we test for whether a
mixture of Gaussians provides a better fit relative to a single Gaussian,
without assuming that either model is correct. The proposed test has a simple
critical value and provides provable error control. One version of our test
provides exact, finite sample control of the type I error. We show how our
tests can be used for hierarchical clustering as well as in a sequential manner
for model selection. We conclude with an extensive simulation study and a
cluster analysis of a gene expression dataset.
| stat |
Associations between park features, park satisfaction and park use in a
multi-ethnic deprived urban area | Parks are increasingly understood to be key community resources for public
health, particularly for ethnic minority and low socioeconomic groups. At the
same time, research suggests parks are underutilised by these groups. In order
to design effective interventions to promote health, the determinants of park
use for these groups must be understood. This study examines the associations
between park features, park satisfaction and park use in a deprived and
ethnically diverse sample in Bradford, UK. 652 women from the Born in Bradford
cohort completed a survey on park satisfaction and park use. Using a
standardised direct observation tool, 44 parks in the area were audited for
present park features. Features assessed were: access, recreational facilities,
amenities, natural features, significant natural features, non-natural
features, incivilities and usability. Size and proximity to the park were also
calculated. Multilevel linear regressions were performed to understand
associations between park features and (1) park satisfaction and (2) park use.
Interactions between park features, ethnicity and socioeconomic status were
explored, and park satisfaction was tested as a mediator in the relationship
between park features and park use. More amenities and greater usability were
associated with increased park satisfaction, while more incivilities were
negatively related to park satisfaction. Incivilities, access and proximity
were also negatively associated with park use. Ethnicity and socioeconomic
status had no moderating role, and there was no evidence for park satisfaction
as a mediator between park features and park use. Results suggest diverse park
features are associated with park satisfaction and park use, but this did not
vary by ethnicity or socioeconomic status. The reduction of incivilities should
be prioritised where the aim is to encourage park satisfaction and park use.
| stat |
Bayesian model selection in additive partial linear models via locally
adaptive splines | We consider a model selection problem for additive partial linear models that
provide a flexible framework allowing both linear and nonlinear additive
components. In practice, it is challenging to determine which additive
components should be excluded from the model and simultaneously determine
whether nonzero nonlinear components can be further simplified to linear
components in the final model. In this paper, we propose a new Bayesian
framework for data-driven model selection by conducting careful model
specification, including the choice of prior distribution and of nonparametric
model for the nonlinear additive components, and propose new sophisticated
computational strategies. Our models, methods, and algorithms are deployed on a
suite of numerical studies and applied to a nutritional epidemiology study. The
numerical results show that the proposed methodology outperforms previously
available methodologies in terms of effective sample sizes of the Markov chain
samplers and the overall misclassification rates, especially in
high-dimensional and large-sample cases.
| stat |
Adversarial Combinatorial Bandits with General Non-linear Reward
Functions | In this paper we study the adversarial combinatorial bandit with a known
non-linear reward function, extending existing work on adversarial linear
combinatorial bandit. {The adversarial combinatorial bandit with general
non-linear reward is an important open problem in bandit literature, and it is
still unclear whether there is a significant gap from the case of linear
reward, stochastic bandit, or semi-bandit feedback.} We show that, with $N$
arms and subsets of $K$ arms being chosen at each of $T$ time periods, the
minimax optimal regret is $\widetilde\Theta_{d}(\sqrt{N^d T})$ if the reward
function is a $d$-degree polynomial with $d< K$, and $\Theta_K(\sqrt{N^K T})$
if the reward function is not a low-degree polynomial. {Both bounds are
significantly different from the bound $O(\sqrt{\mathrm{poly}(N,K)T})$ for the
linear case, which suggests that there is a fundamental gap between the linear
and non-linear reward structures.} Our result also finds applications to
adversarial assortment optimization problem in online recommendation. We show
that in the worst-case of adversarial assortment problem, the optimal algorithm
must treat each individual $\binom{N}{K}$ assortment as independent.
| stat |
A Potts-Mixture Spatiotemporal Joint Model for Combined MEG and EEG Data | We develop a new methodology for determining the location and dynamics of
brain activity from combined magnetoencephalography (MEG) and
electroencephalography (EEG) data. The resulting inverse problem is ill-posed
and is one of the most difficult problems in neuroimaging data analysis. In our
development we propose a solution that combines the data from three different
modalities, MRI, MEG, and EEG, together. We propose a new Bayesian spatial
finite mixture model that builds on the mesostate-space model developed by
Daunizeau and Friston (2007). Our new model incorporates two major extensions:
(i) We combine EEG and MEG data together and formulate a joint model for
dealing with the two modalities simultaneously; (ii) we incorporate the Potts
model to represent the spatial dependence in an allocation process that
partitions the cortical surface into a small number of latent states termed
mesostates. The cortical surface is obtained from MRI. We formulate the new
spatiotemporal model and derive an efficient procedure for simultaneous point
estimation and model selection based on the iterated conditional modes
algorithm combined with local polynomial smoothing. The proposed method results
in a novel estimator for the number of mixture components and is able to select
active brain regions which correspond to active variables in a high-dimensional
dynamic linear model. The methodology is investigated using synthetic data and
simulation studies and then demonstrated on an application examining the neural
response to the perception of scrambled faces. R software implementing the
methodology along with several sample datasets are available at the following
GitHub repository https://github.com/v2south/PottsMix.
| stat |
$\beta^3$-IRT: A New Item Response Model and its Applications | Item Response Theory (IRT) aims to assess latent abilities of respondents
based on the correctness of their answers in aptitude test items with different
difficulty levels. In this paper, we propose the $\beta^3$-IRT model, which
models continuous responses and can generate a much enriched family of Item
Characteristic Curve (ICC). In experiments we applied the proposed model to
data from an online exam platform, and show our model outperforms a more
standard 2PL-ND model on all datasets. Furthermore, we show how to apply
$\beta^3$-IRT to assess the ability of machine learning classifiers. This novel
application results in a new metric for evaluating the quality of the
classifier's probability estimates, based on the inferred difficulty and
discrimination of data instances.
| stat |
Fast and Simple Natural-Gradient Variational Inference with Mixture of
Exponential-family Approximations | Natural-gradient methods enable fast and simple algorithms for variational
inference, but due to computational difficulties, their use is mostly limited
to \emph{minimal} exponential-family (EF) approximations. In this paper, we
extend their application to estimate \emph{structured} approximations such as
mixtures of EF distributions. Such approximations can fit complex, multimodal
posterior distributions and are generally more accurate than unimodal EF
approximations. By using a \emph{minimal conditional-EF} representation of such
approximations, we derive simple natural-gradient updates. Our empirical
results demonstrate a faster convergence of our natural-gradient method
compared to black-box gradient-based methods with reparameterization gradients.
Our work expands the scope of natural gradients for Bayesian inference and
makes them more widely applicable than before.
| stat |
Efficient Bayesian inference for nonlinear state space models with
univariate autoregressive state equation | Latent autoregressive processes are a popular choice to model time varying
parameters. These models can be formulated as nonlinear state space models for
which inference is not straightforward due to the high number of parameters.
Therefore maximum likelihood methods are often infeasible and researchers rely
on alternative techniques, such as Gibbs sampling. But conventional Gibbs
samplers are often tailored to specific situations and suffer from high
autocorrelation among repeated draws. We present a Gibbs sampler for general
nonlinear state space models with an univariate autoregressive state equation.
For this we employ an interweaving strategy and elliptical slice sampling to
exploit the dependence implied by the autoregressive process. Within a
simulation study we demonstrate the efficiency of the proposed sampler for
bivariate dynamic copula models. Further we are interested in modeling the
volatility return relationship. Therefore we use the proposed sampler to
estimate the parameters of stochastic volatility models with skew Student t
errors and the parameters of a novel bivariate dynamic mixture copula model.
This model allows for dynamic asymmetric tail dependence. Comparison to
relevant benchmark models, such as the DCC-GARCH or a Student t copula model,
with respect to predictive accuracy shows the superior performance of the
proposed approach.
| stat |
A Langevinized Ensemble Kalman Filter for Large-Scale Static and Dynamic
Learning | The Ensemble Kalman Filter (EnKF) has achieved great successes in data
assimilation in atmospheric and oceanic sciences, but its failure in
convergence to the right filtering distribution precludes its use for
uncertainty quantification. We reformulate the EnKF under the framework of
Langevin dynamics, which leads to a new particle filtering algorithm, the
so-called Langevinized EnKF. The Langevinized EnKF inherits the
forecast-analysis procedure from the EnKF and the use of mini-batch data from
the stochastic gradient Langevin-type algorithms, which make it scalable with
respect to both the dimension and sample size. We prove that the Langevinized
EnKF converges to the right filtering distribution in Wasserstein distance
under the big data scenario that the dynamic system consists of a large number
of stages and has a large number of samples observed at each stage. We
reformulate the Bayesian inverse problem as a dynamic state estimation problem
based on the techniques of subsampling and Langevin diffusion process. We
illustrate the performance of the Langevinized EnKF using a variety of
examples, including the Lorenz-96 model, high-dimensional variable selection,
Bayesian deep learning, and Long Short Term Memory (LSTM) network learning with
dynamic data.
| stat |
Amortized Bayesian model comparison with evidential deep learning | Comparing competing mathematical models of complex natural processes is a
shared goal among many branches of science. The Bayesian probabilistic
framework offers a principled way to perform model comparison and extract
useful metrics for guiding decisions. However, many interesting models are
intractable with standard Bayesian methods, as they lack a closed-form
likelihood function or the likelihood is computationally too expensive to
evaluate. With this work, we propose a novel method for performing Bayesian
model comparison using specialized deep learning architectures. Our method is
purely simulation-based and circumvents the step of explicitly fitting all
alternative models under consideration to each observed dataset. Moreover, it
requires no hand-crafted summary statistics of the data and is designed to
amortize the cost of simulation over multiple models and observable datasets.
This makes the method particularly effective in scenarios where model fit needs
to be assessed for a large number of datasets, so that per-dataset inference is
practically infeasible.Finally, we propose a novel way to measure epistemic
uncertainty in model comparison problems. We demonstrate the utility of our
method on toy examples and simulated data from non-trivial models from
cognitive science and single-cell neuroscience. We show that our method
achieves excellent results in terms of accuracy, calibration, and efficiency
across the examples considered in this work. We argue that our framework can
enhance and enrich model-based analysis and inference in many fields dealing
with computational models of natural processes. We further argue that the
proposed measure of epistemic uncertainty provides a unique proxy to quantify
absolute evidence even in a framework which assumes that the true
data-generating model is within a finite set of candidate models.
| stat |
Does Regression Approximate the Influence of the Covariates or Just
Measurement Errors? A Model Validity Test | A criterion is proposed for testing hypothesis about the nature of the error
variance in the dependent variable in linear model, which separates correctly
and incorrectly specified models. In the former only measurement errors
determine the variance (i.e., dependent variable is correctly explained by
independent ones, up to measurement errors), while in the latter the model
lacks some independent covariates (or has nonlinear structure). The proposed
MEM-V (Measurement Error Model Validity) test checks the validity of the model
when both dependent and independent covariates are measured with errors. The
criterion has asymptotic character, but numerical simulations outlined
approximate boundaries where estimates make sense. A practical example of the
implementation of the test is discussed in detail; it shows ability of the test
to detect wrong specification even in seemingly perfect models. This type of
relation between measurement errors and model specification has not been
studied before, and the proposed criterion may stimulate future research in
this important area.
| stat |