title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Robust mixture regression with Exponential Power distribution | Assuming an exponential power distribution is one way to deal with outliers
in regression and clustering, which can increase the robustness of the
analysis. Gaussian distribution is a special case of an exponential
distribution. And an exponential power distribution can be viewed as a scale
mixture of normal distributions. Thus, model selection methods developed for
the Gaussian mixture model can be easily extended for the exponential power
mixture model. Moreover, Gaussian mixture models tend to select more components
than exponential power mixture models in real-world cases, which means
exponential power mixture models are easier to interpret. In this paper, We
develop analyses for mixture regression models when the errors are assumed to
follow an exponential power distribution. It will be robust to outliers, and
model selection for it is easy to implement.
| stat |
Calibrated Surrogate Losses for Adversarially Robust Classification | Adversarially robust classification seeks a classifier that is insensitive to
adversarial perturbations of test patterns. This problem is often formulated
via a minimax objective, where the target loss is the worst-case value of the
0-1 loss subject to a bound on the size of perturbation. Recent work has
proposed convex surrogates for the adversarial 0-1 loss, in an effort to make
optimization more tractable. A primary question is that of consistency, that
is, whether minimization of the surrogate risk implies minimization of the
adversarial 0-1 risk. In this work, we analyze this question through the lens
of calibration, which is a pointwise notion of consistency. We show that no
convex surrogate loss is calibrated with respect to the adversarial 0-1 loss
when restricted to the class of linear models. We further introduce a class of
nonconvex losses and offer necessary and sufficient conditions for losses in
this class to be calibrated. We also show that if the underlying distribution
satisfies Massart's noise condition, convex losses can also be calibrated in
the adversarial setting.
| stat |
Data Validation Infrastructure for R | Checking data quality against domain knowledge is a common activity that
pervades statistical analysis from raw data to output. The R package 'validate'
facilitates this task by capturing and applying expert knowledge in the form of
validation rules: logical restrictions on variables, records, or data sets that
should be satisfied before they are considered valid input for further
analysis. In the validate package, validation rules are objects of computation
that can be manipulated, investigated, and confronted with data or versions of
a data set. The results of a confrontation are then available for further
investigation, summarization or visualization. Validation rules can also be
endowed with metadata and documentation and they may be stored or retrieved
from external sources such as text files or tabular formats. This data
validation infrastructure thus allows for systematic, user-defined definition
of data quality requirements that can be reused for various versions of a data
set or by data correction algorithms that are parameterized by validation
rules.
| stat |
Achieving Robustness to Aleatoric Uncertainty with Heteroscedastic
Bayesian Optimisation | Bayesian optimisation is a sample-efficient search methodology that holds
great promise for accelerating drug and materials discovery programs. A
frequently-overlooked modelling consideration in Bayesian optimisation
strategies however, is the representation of heteroscedastic aleatoric
uncertainty. In many practical applications it is desirable to identify inputs
with low aleatoric noise, an example of which might be a material composition
which consistently displays robust properties in response to a noisy
fabrication process. In this paper, we propose a heteroscedastic Bayesian
optimisation scheme capable of representing and minimising aleatoric noise
across the input space. Our scheme employs a heteroscedastic Gaussian process
(GP) surrogate model in conjunction with two straightforward adaptations of
existing acquisition functions. First, we extend the augmented expected
improvement (AEI) heuristic to the heteroscedastic setting and second, we
introduce the aleatoric noise-penalised expected improvement (ANPEI) heuristic.
Both methodologies are capable of penalising aleatoric noise in the suggestions
and yield improved performance relative to homoscedastic Bayesian optimisation
and random sampling on toy problems as well as on two real-world scientific
datasets. Code is available at:
\url{https://github.com/Ryan-Rhys/Heteroscedastic-BO}
| stat |
Robust inference on population indirect causal effects: the generalized
front-door criterion | Standard methods for inference about direct and indirect effects require
stringent no unmeasured confounding assumptions which often fail to hold in
practice, particularly in observational studies. The goal of this paper is to
introduce a new form of indirect effect, the population intervention indirect
effect (PIIE), that can be nonparametrically identified in the presence of an
unmeasured common cause of exposure and outcome. This new type of indirect
effect captures the extent to which the effect of exposure is mediated by an
intermediate variable under an intervention that holds the component of
exposure directly influencing the outcome at its observed value. The PIIE is in
fact the indirect component of the population intervention effect, introduced
by Hubbard and Van der Laan (2008). Interestingly, our identification criterion
generalizes Judea Pearl's front-door criterion as it does not require no direct
effect of exposure not mediated by the intermediate variable. For inference, we
develop both parametric and semiparametric methods, including a novel doubly
robust semiparametric locally efficient estimator, that perform very well in
simulation studies. Finally, the proposed methods are used to measure the
effectiveness of monetary saving recommendations among women enrolled in a
maternal health program in Tanzania.
| stat |
Deepest Voting: a new way of electing | Classical approval voting processes suffer from well-known impossibility
theorems. We consider grading-based voting, where voters give a grade rather
than a preference to each candidate. Numerous voting processes have been
proposed in this framework, such as majority judgment or range voting, showing
interesting properties. This article aims to present a unified framework for
grading-based voting processes. Our statement is that a voting process can be
seen as the finding of the innermost point of a multivariate scatterplot.
Introducing depth functions in voting theory enables to associate each voting
process to a depth notion, and enlarges the scope of procedures. We focus on a
given family of depth functions, the wL p depths, to highlight the behavior of
the depth-based approach.
| stat |
Bayesian Multivariate Nonlinear State Space Copula Models | In this paper we propose a flexible class of multivariate nonlinear
non-Gaussian state space models, based on copulas. More precisely, we assume
that the observation equation and the state equation are defined by copula
families that are not necessarily equal. For each time point, the resulting
model can be described by a C-vine copula truncated after the first tree, where
the root node is represented by the latent state. Inference is performed within
the Bayesian framework, using the Hamiltonian Monte Carlo method, where a
further D-vine truncated after the first tree is used as prior distribution to
capture the temporal dependence in the latent states. Simulation studies show
that the proposed copula-based approach is extremely flexible, since it is able
to describe a wide range of dependence structures and, at the same time, allows
us to deal with missing data. The application to atmospheric pollutant
measurement data shows that our approach is suitable for accurate modeling and
prediction of data dynamics in the presence of missing values. Comparison to a
Gaussian linear state space model and to Bayesian additive regression trees
shows the superior performance of the proposed model with respect to predictive
accuracy.
| stat |
Bayesian inference for treatment effects under nested subsets of
controls | When constructing a model to estimate the causal effect of a treatment, it is
necessary to control for other factors which may have confounding effects.
Because the ignorability assumption is not testable, however, it is usually
unclear which set of controls is appropriate, and effect estimation is
generally sensitive to this choice. A common approach in this case is to fit
several models, each with a different set of controls, but it is difficult to
reconcile inference under the multiple resulting posterior distributions for
the treatment effect. Therefore we propose a two-stage approach to measure the
sensitivity of effect estimation with respect to control specification. In the
first stage, a model is fit with all available controls using a prior carefully
selected to adjust for confounding. In the second stage, posterior
distributions are calculated for the treatment effect under nested sets of
controls by propagating posterior uncertainty in the original model. We
demonstrate how our approach can be used to detect the most significant
confounders in a dataset, and apply it in a sensitivity analysis of an
observational study measuring the effect of legalized abortion on crime rates.
| stat |
Global sensitivity analysis for stochastic simulators based on
generalized lambda surrogate models | Global sensitivity analysis aims at quantifying the impact of input
variability onto the variation of the response of a computational model. It has
been widely applied to deterministic simulators, for which a set of input
parameters has a unique corresponding output value. Stochastic simulators,
however, have intrinsic randomness due to their use of (pseudo)random numbers,
so they give different results when run twice with the same input parameters
but non-common random numbers. Due to this random nature, conventional Sobol'
indices, used in global sensitivity analysis, can be extended to stochastic
simulators in different ways. In this paper, we discuss three possible
extensions and focus on those that depend only on the statistical dependence
between input and output. This choice ignores the detailed data generating
process involving the internal randomness, and can thus be applied to a wider
class of problems. We propose to use the generalized lambda model to emulate
the response distribution of stochastic simulators. Such a surrogate can be
constructed without the need for replications. The proposed method is applied
to three examples including two case studies in finance and epidemiology. The
results confirm the convergence of the approach for estimating the sensitivity
indices even with the presence of strong heteroskedasticity and small
signal-to-noise ratio.
| stat |
Classification via local manifold approximation | Classifiers label data as belonging to one of a set of groups based on input
features. It is challenging to obtain accurate classification performance when
the feature distributions in the different classes are complex, with nonlinear,
overlapping and intersecting supports. This is particularly true when training
data are limited. To address this problem, this article proposes a new type of
classifier based on obtaining a local approximation to the support of the data
within each class in a neighborhood of the feature to be classified, and
assigning the feature to the class having the closest support. This general
algorithm is referred to as LOcal Manifold Approximation (LOMA) classification.
As a simple and theoretically supported special case having excellent
performance in a broad variety of examples, we use spheres for local
approximation, obtaining a SPherical Approximation (SPA) classifier. We
illustrate substantial gains for SPA over competitors on a variety of
challenging simulated and real data examples.
| stat |
DL-PDE: Deep-learning based data-driven discovery of partial
differential equations from discrete and noisy data | In recent years, data-driven methods have been developed to learn dynamical
systems and partial differential equations (PDE). The goal of such work is
discovering unknown physics and the corresponding equations. However, prior to
achieving this goal, major challenges remain to be resolved, including learning
PDE under noisy data and limited discrete data. To overcome these challenges,
in this work, a deep-learning based data-driven method, called DL-PDE, is
developed to discover the governing PDEs of underlying physical processes. The
DL-PDE method combines deep learning via neural networks and data-driven
discovery of PDE via sparse regressions. In the DL-PDE, a neural network is
first trained, and then a large amount of meta-data is generated, and the
required derivatives are calculated by automatic differentiation. Finally, the
form of PDE is discovered by sparse regression. The proposed method is tested
with physical processes, governed by groundwater flow equation,
convection-diffusion equation, Burgers equation and Korteweg-de Vries (KdV)
equation, for proof-of-concept and applications in real-world engineering
settings. The proposed method achieves satisfactory results when data are noisy
and limited.
| stat |
Unsupervised Ground Metric Learning using Wasserstein Eigenvectors | Optimal Transport (OT) defines geometrically meaningful "Wasserstein"
distances, used in machine learning applications to compare probability
distributions. However, a key bottleneck is the design of a "ground" cost which
should be adapted to the task under study. In most cases, supervised metric
learning is not accessible, and one usually resorts to some ad-hoc approach.
Unsupervised metric learning is thus a fundamental problem to enable
data-driven applications of Optimal Transport. In this paper, we propose for
the first time a canonical answer by computing the ground cost as a positive
eigenvector of the function mapping a cost to the pairwise OT distances between
the inputs. This map is homogeneous and monotone, thus framing unsupervised
metric learning as a non-linear Perron-Frobenius problem. We provide criteria
to ensure the existence and uniqueness of this eigenvector. In addition, we
introduce a scalable computational method using entropic regularization, which
- in the large regularization limit - operates a principal component analysis
dimensionality reduction. We showcase this method on synthetic examples and
datasets. Finally, we apply it in the context of biology to the analysis of a
high-throughput single-cell RNA sequencing (scRNAseq) dataset, to improve cell
clustering and infer the relationships between genes in an unsupervised way.
| stat |
Robust probabilistic modeling of photoplethysmography signals with
application to the classification of premature beats | In this paper we propose a robust approach to model photoplethysmography
(PPG) signals. After decomposing the signal into two components, we focus the
analysis on the pulsatile part, related to cardiac information. The goal is to
enable a deeper understanding of the information contained in the pulse shape,
together with that derived from the rhythm. Our approach combines functional
data analysis with a state space representation and guarantees fitting
robustness and flexibility on stationary signals, without imposing a priori
information on the waveform and heart rhythm. With a Bayesian approach, we
learn the distribution of the parameters, used for understanding and monitoring
PPG signals. The model can be used for data compression, for inferring medical
parameters and to understand condition-related waveform characteristics. In
particular, we detail a procedure for the detection of premature contractions
based on the residuals of the fit. This method can handle both atrial and
ventricular premature contractions, and classify the type by only using
information from the model fit.
| stat |
Scale mixture of skew-normal linear mixed models with within-subject
serial dependence | In longitudinal studies, repeated measures are collected over time and hence
they tend to be serially correlated. In this paper we consider an extension of
skew-normal/independent linear mixed models introduced by Lachos et al. (2010),
where the error term has a dependence structure, such as damped exponential
correlation or autoregressive correlation of order p. The proposed model
provides flexibility in capturing the effects of skewness and heavy tails
simultaneously when continuous repeated measures are serially correlated. For
this robust model, we present an efficient EM-type algorithm for computation of
maximum likelihood estimation of parameters and the observed information matrix
is derived analytically to account for standard errors. The methodology is
illustrated through an application to schizophrenia data and several simulation
studies. The proposed algorithm and methods are implemented in the new R
package skewlmm.
| stat |
Local Indicator of Colocation Quotient with a Statistical Significance
Test: Examining Spatial Association of Crime and Facilities | Most existing point-based colocation methods are global measures (e.g., join
count statistic, cross K function, and global colocation quotient). Most
recently, a local indicator such as the local colocation quotient is proposed
to capture the variability of colocation across areas. Our research advances
this line of work by developing a simulation-based statistic test for the local
indicator of colocation quotient (LCLQ). The study applies the indicator to
examine the association of land use facilities with crime patterns. Moreover,
we use the street network distance in addition to the traditional Euclidean
distance in defining neighbors since human activities (including facilities and
crimes) usually occur along a street network. The method is applied to analyze
the colocation of three types of crimes and three categories of facilities in a
city in Jiangsu Province, China. The findings demonstrate the value of the
proposed method in colocation analysis of crime and facilities, and in general
colocation analysis of point data.
| stat |
Joint Latent Class Trees: A Tree-Based Approach to Modeling
Time-to-event and Longitudinal Data | In this paper, we propose a semiparametric, tree based joint latent class
modeling approach (JLCT) to model the joint behavior of longitudinal and
time-to-event data. Existing joint latent class modeling approaches are
parametric and can suffer from high computational cost. The most common
parametric approach, the joint latent class model (JLCM), further restricts
analysis to using time-invariant covariates in modeling survival risks and
latent class memberships. Instead, the proposed JLCT is fast to fit, and can
use time-varying covariates in all of its modeling components. We demonstrate
the prognostic value of using time-varying covariates, and therefore the
advantage of JLCT over JLCM on simulated data. We further apply JLCT to the
PAQUID data set and confirm its superior prediction performance and
orders-of-magnitude speedup over JLCM.
| stat |
Sequential Neural Posterior and Likelihood Approximation | We introduce the sequential neural posterior and likelihood approximation
(SNPLA) algorithm. SNPLA is a normalizing flows-based algorithm for inference
in implicit models. Thus, SNPLA is a simulation-based inference method that
only requires simulations from a generative model. Compared to similar methods,
the main advantage of SNPLA is that our method jointly learns both the
posterior and the likelihood. SNPLA completely avoid Markov chain Monte Carlo
sampling and correction-steps of the parameter proposal function that are
introduced in similar methods, but that can be numerically unstable or
restrictive. Over four experiments, we show that SNPLA performs competitively
when utilizing the same number of model simulations as used in other methods,
even though the inference problem for SNPLA is more complex due to the joint
learning of posterior and likelihood function.
| stat |
Nonparametric tests of independence for circular data based on
trigonometric moments | We introduce nonparametric tests of independence for bivariate circular data
based on trigonometric moments. Our contributions lie in (i) proposing
nonparametric tests that are locally and asymptotically optimal against
bivariate cosine von Mises alternatives and (ii) extending these tests, via the
empirical characteristic function, to obtain consistent tests against broader
sets of alternatives, eventually being omnibus. We thus provide a collection of
trigonometric-based tests of varying generality and known optimalities. The
large-sample behaviours of the tests under the null and alternative hypotheses
are obtained, while simulations show that the new tests are competitive against
previous proposals. Two data applications in astronomy and forest science
illustrate the usage of the tests.
| stat |
The Dinegentropy of Diagnostic Tests | Diagnostic testing is germane to a variety of scenarios in medicine, pandemic
tracking, threat detection, and signal processing. This is an expository paper
with some original results. Here we first set up a mathematical architecture
for diagnostics, and explore its probabilistic underpinnings. Doing so enables
us to develop new metrics for assessing the efficacy of different kinds of
diagnostic tests, and for solving a long standing open problem in diagnostics,
namely, comparing tests when their receiver operating characteristic curves
cross. The first is done by introducing the notion of what we call, a Gini
Coefficient; the second by invoking the information theoretic notion of
dinegentropy. Taken together, these may be seen a contribution to the state of
the art of diagnostics. The spirit of our work could also be relevant to the
much discussed topic of batch testing, where each batch is defined by the
partitioning strategy used to create it. However this possibility has not been
explored here in any detail. Rather, we invite the attention of other
researchers to investigate this idea, as future work.
| stat |
Determining the Dependence Structure of Multivariate Extremes | In multivariate extreme value analysis, the nature of the extremal dependence
between variables should be considered when selecting appropriate statistical
models. Interest often lies with determining which subsets of variables can
take their largest values simultaneously, while the others are of smaller
order. Our approach to this problem exploits hidden regular variation
properties on a collection of non-standard cones and provides a new set of
indices that reveal aspects of the extremal dependence structure not available
through existing measures of dependence. We derive theoretical properties of
these indices, demonstrate their value through a series of examples, and
develop methods of inference that also estimate the proportion of extremal mass
associated with each cone. We apply the methods to UK river flows, estimating
the probabilities of different subsets of sites being large simultaneously.
| stat |
Some examples of application for predicting of compressive sensing
method | This paper considers application of the SALSA algorithm as a method of
forecasting and applies it to simulated electrical signal, temperature
recording from the Australian Bureau of Meteorology and stock prices from the
Australian stock exchange. It compares it to basic linear extrapolation and
casual smoothing extrapolation, in all cases SALSA extrapolation proves to be a
better method of forecasting than linear extrapolation. However, it cannot be
imperially stated that it is superior to Causal smoothing extrapolation in
complex systems as it has a higher L2 euclidean in these experiments. while
usually retaining more shape and statistical elements of the original function
than Causal smoothing extrapolation. Leading to the conclusion the Causal
Smoothing extrapolation can provide a more conservative forecast for complex
systems while the SALSA algorithm more accurately predicts the range of
possible events as well as being the superior forecasting method for electrical
signals, the physical process it is designed to forecast.
| stat |
The Evidence Lower Bound of Variational Autoencoders Converges to a Sum
of Three Entropies | The central objective function of a variational autoencoder (VAE) is its
variational lower bound. Here we show that for standard VAEs the variational
bound is at convergence equal to the sum of three entropies: the (negative)
entropy of the latent distribution, the expected (negative) entropy of the
observable distribution, and the average entropy of the variational
distributions. Our derived analytical results are exact and apply for small as
well as complex neural networks for decoder and encoder. Furthermore, they
apply for finite and infinitely many data points and at any stationary point
(including local and global maxima). As a consequence, we show that the
variance parameters of encoder and decoder play the key role in determining the
values of variational bounds at convergence. Furthermore, the obtained results
can allow for closed-form analytical expressions at convergence, which may be
unexpected as neither variational bounds of VAEs nor log-likelihoods of VAEs
are closed-form during learning. As our main contribution, we provide the
proofs for convergence of standard VAEs to sums of entropies. Furthermore, we
numerically verify our analytical results and discuss some potential
applications. The obtained equality to entropy sums provides novel information
on those points in parameter space that variational learning converges to. As
such, we believe they can potentially significantly contribute to our
understanding of established as well as novel VAE approaches.
| stat |
Probabilistic Forecasting of Regional Net-load with Conditional Extremes
and Gridded NWP | The increasing penetration of embedded renewables makes forecasting net-load,
consumption less embedded generation, a significant and growing challenge. Here
a framework for producing probabilistic forecasts of net-load is proposed with
particular attention given to the tails of predictive distributions, which are
required for managing risk associated with low-probability events. Only small
volumes of data are available in the tails, by definition, so estimation of
predictive models and forecast evaluation requires special attention. We
propose a solution based on a best-in-class load forecasting methodology
adapted for net-load, and model the tails of predictive distributions with the
Generalised Pareto Distribution, allowing its parameters to vary smoothly as
functions of covariates. The resulting forecasts are shown to be calibrated and
sharper than those produced with unconditional tail distributions. In a
use-case inspired evaluation exercise based on reserve setting, the conditional
tails are shown to reduce the overall volume of reserve required to manage a
given risk. Furthermore, they identify periods of high risk not captured by
other methods. The proposed method therefore enables user to both reduce costs
and avoid excess risk.
| stat |
On the Theoretical Properties of the Exchange Algorithm | Exchange algorithm is one of the most popular extensions of
Metropolis-Hastings algorithm to sample from doubly-intractable distributions.
However, theoretical exploration of exchange algorithm is very limited. For
example, natural questions like `Does exchange algorithm converge at a
geometric rate?' or `Does the exchange algorithm admit a Central Limit
Theorem?' have not been answered. In this paper, we study the theoretical
properties of exchange algorithm, in terms of asymptotic variance and
convergence speed. We compare the exchange algorithm with the original
Metropolis-Hastings algorithm and provide both necessary and sufficient
conditions for geometric ergodicity of the exchange algorithm, which can be
applied to various practical applications such as exponential random graph
models and Ising models. A central limit theorem for the exchange algorithm is
also established. Meanwhile, a concrete example, involving the Binomial model
with conjugate and non-conjugate priors, is treated in detail with sharp
convergence rates. Our results justify the theoretical usefulness of the
exchange algorithm.
| stat |
Generalized Single Index Models and Jensen Effects on Reproduction and
Survival | Environmental variability often has substantial impacts on natural
populations and communities through its effects on the performance of
individuals. Because organisms' responses to environmental conditions are often
nonlinear (e.g., decreasing performance on both sides of an optimal
temperature), the mean response is often different from the response in the
mean environment. Ye et. al. 2020, proposed testing for the presence of such
variance effects on individual or population growth rates by estimating the
"Jensen Effect", the difference in average growth rates under varying versus
fixed environments, in functional single index models for environmental effects
on growth. In this paper, we extend this analysis to effect of environmental
variance on reproduction and survival, which have count and binary outcomes. In
the standard generalized linear models used to analyze such data the direction
of the Jensen Effect is tacitly assumed a priori by the model's link function.
Here we extend the methods of Ye et. al. 2020 using a generalized single index
model to test whether this assumed direction is contradicted by the data. We
show that our test has reasonable power under mild alternatives, but requires
sample sizes that are larger than are often available. We demonstrate our
methods on a long-term time series of plant ground cover on the Idaho steppe.
| stat |
Controlling for Unmeasured Confounding in the Presence of Time:
Instrumental Variable for Trend | Unmeasured confounding is a key threat to reliable causal inference based on
observational studies. We propose a new method called instrumental variable for
trend that explicitly leverages exogenous randomness in the exposure trend to
estimate the average and conditional average treatment effect in the presence
of unmeasured confounding. Specifically, we use an instrumental variable for
trend, a variable that (i) is associated with trend in exposure; (ii) is
independent of the potential exposures, potential trends in outcome and
individual treatment effect; and (iii) has no direct effect on the trend in
outcome and does not modify the individual treatment effect. We develop the
identification assumptions using the potential outcomes framework and we
propose two measures of weak identification. In addition, we present a Wald
estimator and a class of multiply robust and efficient semiparametric
estimators, with provable consistency and asymptotic normality. Furthermore, we
propose a two-sample summary-data Wald estimator to facilitate investigations
of delayed treatment effect. We demonstrate our results in simulated and real
datasets.
| stat |
A Sparsity Inducing Nuclear-Norm Estimator (SpINNEr) for Matrix-Variate
Regression in Brain Connectivity Analysis | Classical scalar-response regression methods treat covariates as a vector and
estimate a corresponding vector of regression coefficients. In medical
applications, however, regressors are often in a form of multi-dimensional
arrays. For example, one may be interested in using MRI imaging to identify
which brain regions are associated with a health outcome. Vectorizing the
two-dimensional image arrays is an unsatisfactory approach since it destroys
the inherent spatial structure of the images and can be computationally
challenging. We present an alternative approach - regularized matrix regression
- where the matrix of regression coefficients is defined as a solution to the
specific optimization problem. The method, called SParsity Inducing Nuclear
Norm EstimatoR (SpINNEr), simultaneously imposes two penalty types on the
regression coefficient matrix---the nuclear norm and the lasso norm---to
encourage a low rank matrix solution that also has entry-wise sparsity. A
specific implementation of the alternating direction method of multipliers
(ADMM) is used to build a fast and efficient numerical solver. Our simulations
show that SpINNEr outperforms other methods in estimation accuracy when the
response-related entries (representing the brain's functional connectivity) are
arranged in well-connected communities. SpINNEr is applied to investigate
associations between HIV-related outcomes and functional connectivity in the
human brain.
| stat |
Discussion of "A Gibbs sampler for a class of random convex polytopes" | An exciting new algorithmic breakthrough has been advanced for how to carry
out inferences in a Dempster-Shafer (DS) formulation of a categorical data
generating model. The developed sampling mechanism, which draws on theory for
directed graphs, is a clever and remarkable achievement, as this has been an
open problem for many decades. In this discussion, I comment on important
contributions, central questions, and prevailing matters of the article.
| stat |
A parsimonious family of multivariate Poisson-lognormal distributions
for clustering multivariate count data | Multivariate count data are commonly encountered through high-throughput
sequencing technologies in bioinformatics, text mining, or in sports analytics.
Although the Poisson distribution seems a natural fit to these count data, its
multivariate extension is computationally expensive.In most cases mutual
independence among the variables is assumed, however this fails to take into
account the correlation among the variables usually observed in the data.
Recently, mixtures of multivariate Poisson-lognormal (MPLN) models have been
used to analyze such multivariate count measurements with a dependence
structure. In the MPLN model, each count is modeled using an independent
Poisson distribution conditional on a latent multivariate Gaussian variable.
Due to this hierarchical structure, the MPLN model can account for
over-dispersion as opposed to the traditional Poisson distribution and allows
for correlation between the variables. Rather than relying on a Monte
Carlo-based estimation framework which is computationally inefficient, a fast
variational-EM based framework is used here for parameter estimation. Further,
a parsimonious family of mixtures of Poisson-lognormal distributions are
proposed by decomposing the covariance matrix and imposing constraints on these
decompositions. Utility of such models is shown using simulated and benchmark
datasets.
| stat |
Cross-study learning for generalist and specialist predictions | The integration and use of data from multiple studies, for the development of
prediction models is an important task in several scientific fields. We propose
a framework for generalist and specialist predictions that leverages multiple
datasets, with potential differences in the relationships between predictors
and outcomes. Our framework uses stacking, and it includes three major
components: 1) an ensemble of prediction models trained on one or more
datasets, 2) task-specific utility functions and 3) a no-data-reuse technique
for estimating stacking weights. We illustrate that under mild regularity
conditions the framework produces stacked prediction functions with oracle
properties. In particular we show that the the stacking weights are nearly
optimal. We also provide sufficient conditions under which the proposed
no-data-reuse technique increases prediction accuracy compared to stacking with
data reuse. We perform a simulation study to illustrate these results. We apply
our framework to predict mortality using a collection of datasets on long-term
exposure to air pollutants.
| stat |
Interpretable Biomanufacturing Process Risk and Sensitivity Analyses for
Quality-by-Design and Stability Control | While biomanufacturing plays a significant role in supporting the economy and
ensuring public health, it faces critical challenges, including complexity,
high variability, lengthy lead time, and very limited process data, especially
for personalized new cell and gene biotherapeutics. Driven by these challenges,
we propose an interpretable semantic bioprocess probabilistic knowledge graph
and develop a game theory based risk and sensitivity analyses for production
process to facilitate quality-by-design and stability control. Specifically, by
exploring the causal relationships and interactions of critical process
parameters and quality attributes (CPPs/CQAs), we create a Bayesian network
based probabilistic knowledge graph characterizing the complex causal
interdependencies of all factors. Then, we introduce a Shapley value based
sensitivity analysis, which can correctly quantify the variation contribution
from each input factor on the outputs (i.e., productivity, product quality).
Since the bioprocess model coefficients are learned from limited process
observations, we derive the Bayesian posterior distribution to quantify model
uncertainty and further develop the Shapley value based sensitivity analysis to
evaluate the impact of estimation uncertainty from each set of model
coefficients. Therefore, the proposed bioprocess risk and sensitivity analyses
can identify the bottlenecks, guide the reliable process specifications and the
most "informative" data collection, and improve production stability.
| stat |
UEFA against the champions? An evaluation of the recent reform of the
Champions League qualification | The UEFA Champions League is the major European club football competition
organised by the Union of European Football Associations (UEFA). It contains 32
teams, hence the national champions of most UEFA associations have to play in
the qualification to receive a slot in its group stage. The paper evaluates the
impact of reforming the Champions Path of the qualifying system, effective from
the 2018/19 season. While it is anticipated that the reduction in the number of
berths decreases the probability of advancing to the group stage, the
distribution of the losses among the national associations can only be
estimated via Monte-Carlo simulations. In contrast to previous studies, our
methodology considers five seasons instead of one to filter out any possible
season-specific attributes. Almost all of the 45 countries are found to gain
less prize money on average. Several champions, including the Cypriot, the
Swiss, and the Scottish, might face a loss of over one million Euros. Since the
negative effects depend to a large extent on the somewhat arbitrary differences
between the positions of the access list, we propose to introduce some
randomness into the determination of entry stages.
| stat |
Robust Mean Estimation in High Dimensions via $\ell_0$ Minimization | We study the robust mean estimation problem in high dimensions, where $\alpha
<0.5$ fraction of the data points can be arbitrarily corrupted. Motivated by
compressive sensing, we formulate the robust mean estimation problem as the
minimization of the $\ell_0$-`norm' of the outlier indicator vector, under
second moment constraints on the inlier data points. We prove that the global
minimum of this objective is order optimal for the robust mean estimation
problem, and we propose a general framework for minimizing the objective. We
further leverage the $\ell_1$ and $\ell_p$ $(0<p<1)$, minimization techniques
in compressive sensing to provide computationally tractable solutions to the
$\ell_0$ minimization problem. Both synthetic and real data experiments
demonstrate that the proposed algorithms significantly outperform
state-of-the-art robust mean estimation methods.
| stat |
Spillover Effects in the Presence of Unobserved Networks | When experimental subjects can interact with each other, the outcome of one
individual may be affected by the treatment status of others. In many social
science experiments, such spillover effects may occur through multiple
networks, for example, through both online and offline face-to-face networks in
a Twitter experiment. Thus, to understand how people use different networks, it
is essential to estimate the spillover effect in each specific network
separately. However, the unbiased estimation of these network-specific
spillover effects requires an often-violated assumption that researchers
observe all relevant networks. We show that, unlike conventional omitted
variable bias, bias due to unobserved networks remains even when treatment
assignment is randomized and when unobserved networks and a network of interest
are independently generated. We then develop parametric and nonparametric
sensitivity analysis methods, with which researchers can assess the potential
influence of unobserved networks on causal findings. We illustrate the proposed
methods with a simulation study based on a real-world Twitter network and an
empirical application based on a network field experiment in China.
| stat |
Evaluating the Success of a Data Analysis | A fundamental problem in the practice and teaching of data science is how to
evaluate the quality of a given data analysis, which is different than the
evaluation of the science or question underlying the data analysis. Previously,
we defined a set of principles for describing data analyses that can be used to
create a data analysis and to characterize the variation between data analyses.
Here, we introduce a metric of quality evaluation that we call the success of a
data analysis, which is different than other potential metrics such as
completeness, validity, or honesty. We define a successful data analysis as the
matching of principles between the analyst and the audience on which the
analysis is developed. In this paper, we propose a statistical model and
general framework for evaluating the success of a data analysis. We argue that
this framework can be used as a guide for practicing data scientists and
students in data science courses for how to build a successful data analysis.
| stat |
RefCurv: A Software for the Construction of Pediatric Reference Curves | In medicine, reference curves serve as an important tool for everyday
clinical practice. Pediatricians assess the growth process of children with the
help of percentile curves serving as norm references. The mathematical methods
for the construction of these reference curves are sophisticated and often
require technical knowledge beyond the scope of physicians. An easy-to-use
software for life scientists and physicians is missing. As a consequence, most
medical publications do not document the construction properly. This project
aims to develop a software that enables non-technical users to apply modern
statistical methods to create and analyze reference curves. In this paper, we
present RefCurv, a software that facilitates the construction of reference
curves. The software comprises functionalities to select and visualize data.
Users can fit models to the data and graphically present them as percentile
curves. Furthermore, the software provides features to highlight possible
outliers, perform model selection, and analyze the sensitivity. RefCurv is an
open-source software with a graphical user interface (GUI) written in Python.
It uses R and the gamlss add-on package (Rigby and Stasinopoulos (2005)) as the
underlying statistical engine. In summary, RefCurv is the first software based
on the gamlss package, which enables practitioners to construct and analyze
reference curves in a user-friendly GUI. In broader terms, the software brings
together the fields of statistical learning and medical application.
Consequently, RefCurv can help to establish the construction of reference
curves in other medical fields.
| stat |
Percentile-Based Residuals for Model Assessment | Residuals are a key component of diagnosing model fit. The usual practice is
to compute standardized residuals using expected values and standard deviations
of the observed data, then use these values to detect outliers and assess model
fit. Approximate normality of these residuals is key for this process to have
good properties, but in many modeling contexts, especially for complex,
multi-level models, normality may not hold. In these cases outlier detection
and model diagnostics aren't properly calibrated. Alternatively, as we
demonstrate, residuals computed from the percentile location of a datum's value
in its full predictive distribution lead to well calibrated evaluations of
model fit. We generalize an approach described by Dunn and Smyth (1996) and
evaluate properties mathematically, via case-studies and by simulation. In
addition, we show that the standard residuals can be calibrated to mimic the
percentile approach, but that this extra step is avoided by directly using
percentile-based residuals. For both the percentile-based residuals and the
calibrated standard residuals, the use of full predictive distributions with
the appropriate location, spread and shape is necessary for valid assessments.
| stat |
Identification of causal effects in case-control studies | Case-control designs are an important tool in contrasting the effects of
well-defined treatments. In this paper, we reconsider classical concepts,
assumptions and principles and explore when the results of case-control studies
can be endowed a causal interpretation. Our focus is on identification of
target causal quantities, or estimands. We cover various estimands relating to
intention-to-treat or per-protocol effects for popular sampling schemes
(case-base, survivor, and risk-set sampling), each with and without matching.
Our approach may inform future research on different estimands, other
variations of the case-control design or settings with additional complexities.
| stat |
Social Media Integration of Flood Data: A Vine Copula-Based Approach | Floods are the most common and among the most severe natural disasters in
many countries around the world. As global warming continues to exacerbate sea
level rise and extreme weather, governmental authorities and environmental
agencies are facing the pressing need of timely and accurate evaluations and
predictions of flood risks. Current flood forecasts are generally based on
historical measurements of environmental variables at monitoring stations. In
recent years, in addition to traditional data sources, large amounts of
information related to floods have been made available via social media.
Members of the public are constantly and promptly posting information and
updates on local environmental phenomena on social media platforms. Despite the
growing interest of scholars towards the usage of online data during natural
disasters, the majority of studies focus exclusively on social media as a
stand-alone data source, while its joint use with other type of information is
still unexplored. In this paper we propose to fill this gap by integrating
traditional historical information on floods with data extracted by Twitter and
Google Trends. Our methodology is based on vine copulas, that allow us to
capture the dependence structure among the marginals, which are modelled via
appropriate time series methods, in a very flexible way. We apply our
methodology to data related to three different coastal locations in the South
cost of the UK. The results show that our approach, based on the integration of
social media data, outperforms traditional methods, providing a more accurate
evaluation and prediction of flood events.
| stat |
Ground Metric Learning on Graphs | Optimal transport (OT) distances between probability distributions are
parameterized by the ground metric they use between observations. Their
relevance for real-life applications strongly hinges on whether that ground
metric parameter is suitably chosen. Selecting it adaptively and
algorithmically from prior knowledge, the so-called ground metric learning GML)
problem, has therefore appeared in various settings. We consider it in this
paper when the learned metric is constrained to be a geodesic distance on a
graph that supports the measures of interest. This imposes a rich structure for
candidate metrics, but also enables far more efficient learning procedures when
compared to a direct optimization over the space of all metric matrices. We use
this setting to tackle an inverse problem stemming from the observation of a
density evolving with time: we seek a graph ground metric such that the OT
interpolation between the starting and ending densities that result from that
ground metric agrees with the observed evolution. This OT dynamic framework is
relevant to model natural phenomena exhibiting displacements of mass, such as
for instance the evolution of the color palette induced by the modification of
lighting and materials.
| stat |
Regularized Maximum Likelihood Estimation for the Random Coefficients
Model | The random coefficients model $Y_i={\beta_0}_i+{\beta_1}_i
{X_1}_i+{\beta_2}_i {X_2}_i+\ldots+{\beta_d}_i {X_d}_i$, with $\mathbf{X}_i$,
$Y_i$, $\mathbf{\beta}_i$ i.i.d, and $\mathbf{\beta}_i$ independent of $X_i$ is
often used to capture unobserved heterogeneity in a population. We propose a
quasi-maximum likelihood method to estimate the joint density distribution of
the random coefficient model. This method implicitly involves the inversion of
the Radon transformation in order to reconstruct the joint distribution, and
hence is an inverse problem. Nonparametric estimation for the joint density of
$\mathbf{\beta}_i=({\beta_0}_i,\ldots, {\beta_d}_i)$ based on kernel methods or
Fourier inversion have been proposed in recent years. Most of these methods
assume a heavy tailed design density $f_\mathbf{X}$. To add stability to the
solution, we apply regularization methods. We analyze the convergence of the
method without assuming heavy tails for $f_\mathbf{X}$ and illustrate
performance by applying the method on simulated and real data. To add stability
to the solution, we apply a Tikhonov-type regularization method.
| stat |
A linear algorithm for multi-target tracking in the context of
possibility theory | We present a modelling framework for multi-target tracking based on
possibility theory and illustrate its ability to account for the general lack
of knowledge that the target-tracking practitioner must deal with when working
with real data. We also introduce and study variants of the notions of point
process and intensity function, which lead to the derivation of an analogue of
the probability hypothesis density (PHD) filter. The gains provided by the
considered modelling framework in terms of flexibility lead to the loss of some
of the abilities that the PHD filter possesses; in particular the estimation of
the number of targets by integration of the intensity function. Yet, the
proposed recursion displays a number of advantages such as facilitating the
introduction of observation-driven birth schemes and the modelling the absence
of information on the initial number of targets in the scene. The performance
of the proposed approach is demonstrated on simulated data.
| stat |
A Bayesian cohort component projection model to estimate adult
populations at the subnational level in data-sparse settings | Accurate estimates of subnational populations are important for policy
formulation and monitoring population health indicators. For example, estimates
of the number of women of reproductive age are important to understand the
population at risk to maternal mortality and unmet need for contraception.
However, in many low-income countries, data on population counts and components
of population change are limited, and so levels and trends subnationally are
unclear. We present a Bayesian constrained cohort component model for the
estimation and projection of subnational populations. The model builds on a
cohort component projection framework, incorporates census data and estimates
from the United Nation's World Population Prospects, and uses characteristic
mortality schedules to obtain estimates of population counts and the components
of population change, including internal migration. The data required as inputs
to the model are minimal and available across a wide range of countries,
including most low-income countries. The model is applied to estimate and
project populations by county in Kenya for 1979-2019, and validated against the
2019 Kenyan census.
| stat |
ENTMOOT: A Framework for Optimization over Ensemble Tree Models | Gradient boosted trees and other regression tree models perform well in a
wide range of real-world, industrial applications. These tree models (i) offer
insight into important prediction features, (ii) effectively manage sparse
data, and (iii) have excellent prediction capabilities. Despite their
advantages, they are generally unpopular for decision-making tasks and
black-box optimization, which is due to their difficult-to optimize structure
and the lack of a reliable uncertainty measure. ENTMOOT is our new framework
for integrating (already trained) tree models into larger optimization
problems. The contributions of ENTMOOT include: (i) explicitly introducing a
reliable uncertainty measure that is compatible with tree models, (ii) solving
the larger optimization problems that incorporate these uncertainty aware tree
models, (iii) proving that the solutions are globally optimal, i.e. no better
solution exists. In particular, we show how the ENTMOOT approach allows a
simple integration of tree models into decision-making and black-box
optimization, where it proves as a strong competitor to commonly-used
frameworks.
| stat |
Stochastic algorithms with descent guarantees for ICA | Independent component analysis (ICA) is a widespread data exploration
technique, where observed signals are modeled as linear mixtures of independent
components. From a machine learning point of view, it amounts to a matrix
factorization problem with a statistical independence criterion. Infomax is one
of the most used ICA algorithms. It is based on a loss function which is a
non-convex log-likelihood. We develop a new majorization-minimization framework
adapted to this loss function. We derive an online algorithm for the streaming
setting, and an incremental algorithm for the finite sum setting, with the
following benefits. First, unlike most algorithms found in the literature, the
proposed methods do not rely on any critical hyper-parameter like a step size,
nor do they require a line-search technique. Second, the algorithm for the
finite sum setting, although stochastic, guarantees a decrease of the loss
function at each iteration. Experiments demonstrate progress on the
state-of-the-art for large scale datasets, without the necessity for any manual
parameter tuning.
| stat |
Stop the Clock: Are Timeout Effects Real? | Timeout is a short interruption during games used to communicate a change in
strategy, to give the players a rest or to stop a negative flow in the game.
Whatever the reason, coaches expect an improvement in their team's performance
after a timeout. But how effective are these timeouts in doing so? The simple
average of the differences between the scores before and after the timeouts has
been used as evidence that there is an effect and that it is substantial. We
claim that these statistical averages are not proper evidence and a more sound
approach is needed. We applied a formal causal framework using a large dataset
of official NBA play-by-play tables and drew our assumptions about the data
generation process in a causal graph. Using different matching techniques to
estimate the causal effect of timeouts, we concluded that timeouts have no
effect on teams' performances. Actually, since most timeouts are called when
the opposing team is scoring more frequently, the moments that follow resemble
an improvement in the team's performance but are just the natural game tendency
to return to its average state. This is another example of what statisticians
call the regression to the mean phenomenon.
| stat |
Small area estimation of general finite-population parameters based on
grouped data | This paper proposes a new model-based approach to small area estimation of
general finite-population parameters based on grouped data or frequency data,
which is often available from sample surveys. Grouped data contains information
on frequencies of some pre-specified groups in each area, for example the
numbers of households in the income classes, and thus provides more detailed
insight about small areas than area-level aggregated data. A direct application
of the widely used small area methods, such as the Fay-Herriot model for
area-level data and nested error regression model for unit-level data, is not
appropriate since they are not designed for grouped data. The newly proposed
method adopts the multinomial likelihood function for the grouped data. In
order to connect the group probabilities of the multinomial likelihood and the
auxiliary variables within the framework of small area estimation, we introduce
the unobserved unit-level quantities of interest which follows the linear mixed
model with the random intercepts and dispersions after some transformation.
Then the probabilities that a unit belongs to the groups can be derived and are
used to construct the likelihood function for the grouped data given the random
effects. The unknown model parameters (hyperparameters) are estimated by a
newly developed Monte Carlo EM algorithm using an efficient importance
sampling. The empirical best predicts (empirical Bayes estimates) of small area
parameters can be calculated by a simple Gibbs sampling algorithm. The
numerical performance of the proposed method is illustrated based on the
model-based and design-based simulations. In the application to the city level
grouped income data of Japan, we complete the patchy maps of the Gini
coefficient as well as mean income across the country.
| stat |
AdaPtive Noisy Data Augmentation (PANDA) for Simultaneous Construction
of Multiple Graph Models | We extend the data augmentation technique PANDA by Li et al. (2018) that
regularizes single graph estimation to jointly learning multiple graphical
models with various node types in a unified framework. We design two types of
noise to augment the observed data: the first type regularizes the estimation
of each graph while the second type promotes either the structural similarity,
referred as the \joint group lasso regularization, or the numerical similarity,
referred as the joint fused ridge regularization, among the edges in the same
position across graphs. The computation in PANDA is straightforward and only
involves obtaining maximum likelihood estimator in generalized linear models in
an iterative manner. The simulation studies demonstrate PANDA is non-inferior
to existing joint estimation approaches for Gaussian graphical models, and
significantly improves over the naive differencing approach for non-Gaussian
graphical models. We apply PANDA to a real-life lung cancer microarray data to
simultaneously construct four protein networks.
| stat |
Efficient calibration for high-dimensional computer model output using
basis methods | Calibration of expensive computer models with high-dimensional output fields
can be approached via history matching. If the entire output field is matched,
with patterns or correlations between locations or time points represented,
calculating the distance metric between observational data and model output for
a single input setting requires a time intensive inversion of a
high-dimensional matrix. By using a low-dimensional basis representation rather
than emulating each output individually, we define a metric in the reduced
space that allows the implausibility for the field to be calculated
efficiently, with only small matrix inversions required, using projection that
is consistent with the variance specifications in the implausibility. We show
that projection using the $L_2$ norm can result in different conclusions, with
the ordering of points not maintained on the basis, with implications for both
history matching and probabilistic methods. We demonstrate the scalability of
our method through history matching of the Canadian atmosphere model, CanAM4,
comparing basis methods to emulation of each output individually, showing that
the basis approach can be more accurate, whilst also being more efficient.
| stat |
Inferring median survival differences in general factorial designs via
permutation tests | Factorial survival designs with right-censored observations are commonly
inferred by Cox regression and explained by means of hazard ratios. However, in
case of non-proportional hazards, their interpretation can become cumbersome;
especially for clinicians. We therefore offer an alternative: median survival
times are used to estimate treatment and interaction effects and null
hypotheses are formulated in contrasts of their population versions.
Permutation-based tests and confidence regions are proposed and shown to be
asymptotically valid. Their type-1 error control and power behavior are
investigated in extensive simulations, showing the new methods' wide
applicability. The latter is complemented by an illustrative data analysis.
| stat |
Discovering Latent Covariance Structures for Multiple Time Series | Analyzing multivariate time series data is important to predict future events
and changes of complex systems in finance, manufacturing, and administrative
decisions. The expressiveness power of Gaussian Process (GP) regression methods
has been significantly improved by compositional covariance structures. In this
paper, we present a new GP model which naturally handles multiple time series
by placing an Indian Buffet Process (IBP) prior on the presence of shared
kernels. Our selective covariance structure decomposition allows exploiting
shared parameters over a set of multiple, selected time series. We also
investigate the well-definedness of the models when infinite latent components
are introduced. We present a pragmatic search algorithm which explores a larger
structure space efficiently. Experiments conducted on five real-world data sets
demonstrate that our new model outperforms existing methods in term of
structure discoveries and predictive performances.
| stat |
Causal Inference with Deep Causal Graphs | Parametric causal modelling techniques rarely provide functionality for
counterfactual estimation, often at the expense of modelling complexity. Since
causal estimations depend on the family of functions used to model the data,
simplistic models could entail imprecise characterizations of the generative
mechanism, and, consequently, unreliable results. This limits their
applicability to real-life datasets, with non-linear relationships and high
interaction between variables. We propose Deep Causal Graphs, an abstract
specification of the required functionality for a neural network to model
causal distributions, and provide a model that satisfies this contract:
Normalizing Causal Flows. We demonstrate its expressive power in modelling
complex interactions and showcase applications of the method to machine
learning explainability and fairness, using true causal counterfactuals.
| stat |
Latent variable models for multivariate dyadic data with zero inflation:
Analysis of intergenerational exchanges of family support | Understanding the help and support that is exchanged between family members
of different generations is of increasing importance, with research questions
in sociology and social policy focusing on both predictors of the levels of
help given and received, and on reciprocity between them. We propose general
latent variable models for analysing such data, when helping tendencies in each
direction are measured by multiple binary indicators of specific types of help.
The model combines two continuous latent variables, which represent the helping
tendencies, with two binary latent class variables which allow for high
proportions of responses where no help of any kind is given or received. This
defines a multivariate version of a zero inflation model. The main part of the
models is estimated using MCMC methods, with a bespoke data augmentation
algorithm. We apply the models to analyse exchanges of help between adult
individuals and their non-coresident parents, using survey data from the UK
Household Longitudinal Study.
| stat |
Detecting direct causality in multivariate time series: A comparative
study | The concept of Granger causality is increasingly being applied for the
characterization of directional interactions in different applications. A
multivariate framework for estimating Granger causality is essential in order
to account for all the available information from multivariate time series.
However, the inclusion of non-informative or non-significant variables creates
estimation problems related to the 'curse of dimensionality'. To deal with this
issue, direct causality measures using variable selection and dimension
reduction techniques have been introduced. In this comparative work, the
performance of an ensemble of bivariate and multivariate causality measures in
the time domain is assessed, focusing on dimension reduction causality
measures. In particular, different types of high-dimensional coupled discrete
systems are used (involving up to 100 variables) and the robustness of the
causality measures to time series length and different noise types is examined.
The results of the simulation study highlight the superiority of the dimension
reduction measures, especially for high-dimensional systems.
| stat |
Seeded Binary Segmentation: A general methodology for fast and optimal
change point detection | In recent years, there has been an increasing demand on efficient algorithms
for large scale change point detection problems. To this end, we propose seeded
binary segmentation, an approach relying on a deterministic construction of
background intervals, called seeded intervals, in which single change points
are searched. The final selection of change points based on the candidates from
seeded intervals can be done in various ways, adapted to the problem at hand.
Thus, seeded binary segmentation is easy to adapt to a wide range of change
point detection problems, let that be univariate, multivariate or even
high-dimensional.
We consider the univariate Gaussian change in mean setup in detail. For this
specific case we show that seeded binary segmentation leads to a near-linear
time approach (i.e. linear up to a logarithmic factor) independent of the
underlying number of change points. Furthermore, using appropriate selection
methods, the methodology is shown to be asymptotically minimax optimal. While
computationally more efficient, the finite sample estimation performance
remains competitive compared to state of the art procedures. Moreover, we
illustrate the methodology for high-dimensional settings with an inverse
covariance change point detection problem where our proposal leads to massive
computational gains while still exhibiting good statistical performance.
| stat |
Retiree mortality forecasting: A partial age-range or a full age-range
model? | An essential input of annuity pricing is the future retiree mortality. From
observed age-specific mortality data, modeling and forecasting can be taken
place in two routes. On the one hand, we can first truncate the available data
to retiree ages and then produce mortality forecasts based on a partial
age-range model. On the other hand, with all available data, we can first apply
a full age-range model to produce forecasts and then truncate the mortality
forecasts to retiree ages. We investigate the difference in modeling the
logarithmic transformation of the central mortality rates between a partial
age-range and a full age-range model, using data from mainly developed
countries in the Human Mortality Database (2020). By evaluating and comparing
the short-term point and interval forecast accuracies, we recommend the first
strategy by truncating all available data to retiree ages and then produce
mortality forecasts. However, when considering the long-term forecasts, it is
unclear which strategy is better since it is more difficult to find a model and
parameters that are optimal. This is a disadvantage of using methods based on
time series extrapolation for long-term forecasting. Instead, an expectation
approach, in which experts set a future target, could be considered, noting
that this method has also had limited success in the past.
| stat |
Flow Rate Estimation From Probe Vehicle Data And Sample Size
Requirements | The interest to use probe vehicles for traffic monitoring is growing. This
paper is focused on the estimation of flow rate from probe vehicle data and the
evaluation of sample size requirements. Three cases are considered depending on
the available information on the percentage of probes p and the flow rate
\lambda: i) p is known but \lambda is unknown ii) \lambda is known but p is
unknown, and iii) both parameters are not known. Estimation methods for all
three cases are presented along with the reliability of these estimates. For
the first two cases, count data provide sufficient information to estimate the
unknown parameters p and \lambda individually when the other one is known. For
these two cases, simple analytical expressions are derived to analyze the
accuracy and sample size requirements. However, when both parameters are
unknown then additional information beyond count data is required. The position
of probe vehicles in a queue at signalized intersections is used as additional
information in that case. The results show for how long data need to be
collected to estimate the parameters at acceptable confidence levels.
| stat |
Percent Change Estimation in Large Scale Online Experiments | Online experiments are a fundamental component of the development of
web-facing products. Given their large user-bases, even small product
improvements can have a large impact on user engagement or profits on an
absolute scale. As a result, accurately estimating the relative impact of these
changes is extremely important. I propose an approach based on an objective
Bayesian model to improve the sensitivity of percent change estimation in A/B
experiments. Leveraging pre-period information, this approach produces more
robust and accurate point estimates and up to 50% tighter credible intervals
than traditional methods. The R package abpackage provides an implementation of
the approach.
| stat |
On Symmetric Losses for Learning from Corrupted Labels | This paper aims to provide a better understanding of a symmetric loss. First,
we emphasize that using a symmetric loss is advantageous in the balanced error
rate (BER) minimization and area under the receiver operating characteristic
curve (AUC) maximization from corrupted labels. Second, we prove general
theoretical properties of symmetric losses, including a
classification-calibration condition, excess risk bound, conditional risk
minimizer, and AUC-consistency condition. Third, since all nonnegative
symmetric losses are non-convex, we propose a convex barrier hinge loss that
benefits significantly from the symmetric condition, although it is not
symmetric everywhere. Finally, we conduct experiments to validate the relevance
of the symmetric condition.
| stat |
A statistical theory of semi-supervised learning | We currently lack a solid statistical understanding of semi-supervised
learning methods, instead treating them as a collection of highly effective
tricks. This precludes the principled combination e.g. of Bayesian methods and
semi-supervised learning, as semi-supervised learning objectives are not
currently formulated as likelihoods for an underlying generative model of the
data. Here, we note that standard image benchmark datasets such as CIFAR-10 are
carefully curated, and we provide a generative model describing the curation
process. Under this generative model, several state-of-the-art semi-supervised
learning techniques, including entropy minimization, pseudo-labelling and the
FixMatch family emerge naturally as variational lower-bounds on the
log-likelihood.
| stat |
Modern Multiple Imputation with Functional Data | This work considers the problem of fitting functional models with sparsely
and irregularly sampled functional data. It overcomes the limitations of the
state-of-the-art methods, which face major challenges in the fitting of more
complex non-linear models. Currently, many of these models cannot be
consistently estimated unless the number of observed points per curve grows
sufficiently quickly with the sample size, whereas, we show numerically that a
modified approach with more modern multiple imputation methods can produce
better estimates in general. We also propose a new imputation approach that
combines the ideas of {\it MissForest} with {\it Local Linear Forest} and
compare their performance with {\it PACE} and several other multivariate
multiple imputation methods. This work is motivated by a longitudinal study on
smoking cessation, in which the Electronic Health Records (EHR) from Penn State
PaTH to Health allow for the collection of a great deal of data, with highly
variable sampling. To illustrate our approach, we explore the relation between
relapse and diastolic blood pressure. We also consider a variety of simulation
schemes with varying levels of sparsity to validate our methods.
| stat |
Disentangling and Learning Robust Representations with Natural
Clustering | Learning representations that disentangle the underlying factors of
variability in data is an intuitive way to achieve generalization in deep
models. In this work, we address the scenario where generative factors present
a multimodal distribution due to the existence of class distinction in the
data. We propose N-VAE, a model which is capable of separating factors of
variation which are exclusive to certain classes from factors that are shared
among classes. This model implements an explicitly compositional latent
variable structure by defining a class-conditioned latent space and a shared
latent space. We show its usefulness for detecting and disentangling
class-dependent generative factors as well as its capacity to generate
artificial samples which contain characteristics unseen in the training data.
| stat |
A robust and efficient algorithm to find profile likelihood confidence
intervals | Profile likelihood confidence intervals are a robust alternative to Wald's
method if the asymptotic properties of the maximum likelihood estimator are not
met. However, the constrained optimization problem defining profile likelihood
confidence intervals can be difficult to solve in these situations, because the
likelihood function may exhibit unfavorable properties. As a result, existing
methods may be inefficient and yield misleading results. In this paper, we
address this problem by computing profile likelihood confidence intervals via a
trust-region approach, where steps computed based on local approximations are
constrained to regions where these approximations are sufficiently precise. As
our algorithm also accounts for numerical issues arising if the likelihood
function is strongly non-linear or parameters are not estimable, the method is
applicable in many scenarios where earlier approaches are shown to be
unreliable. To demonstrate its potential in applications, we apply our
algorithm to benchmark problems and compare it with 6 existing approaches to
compute profile likelihood confidence intervals. Our algorithm consistently
achieved higher success rates than any competitor while also being among the
quickest methods. As our algorithm can be applied to compute both confidence
intervals of parameters and model predictions, it is useful in a wide range of
scenarios.
| stat |
Low-cost sensor networks and land-use regression: interpolating nitrogen
dioxide concentration at high temporal and spatial resolution in Southern
California | The development of low-cost sensors and novel calibration algorithms offer
new opportunities to supplement existing regulatory networks to measure air
pollutants at a high spatial resolution and at hourly and sub-hourly
timescales. We use a random forest model on data from a network of low-cost
sensors to describe the effect of land use features on local-scale air quality,
extend this model to describe the hourly-scale variation of air quality at high
spatial resolution, and show that deviations from the model can be used to
identify particular conditions and locations where air quality differs from the
expected land-use effect. The conditions and locations under which deviations
were detected conform to expectations based on general experience.
| stat |
The flare Package for High Dimensional Linear Regression and Precision
Matrix Estimation in R | This paper describes an R package named flare, which implements a family of
new high dimensional regression methods (LAD Lasso, SQRT Lasso, $\ell_q$ Lasso,
and Dantzig selector) and their extensions to sparse precision matrix
estimation (TIGER and CLIME). These methods exploit different nonsmooth loss
functions to gain modeling flexibility, estimation robustness, and tuning
insensitiveness. The developed solver is based on the alternating direction
method of multipliers (ADMM). The package flare is coded in double precision C,
and called from R by a user-friendly interface. The memory usage is optimized
by using the sparse matrix output. The experiments show that flare is efficient
and can scale up to large problems.
| stat |
Instrumental variable estimation of early treatment effect in randomized
screening trials | The primary analysis of randomized cancer screening trials for cancer
typically adheres to the intention-to-screen principle, measuring
cancer-specific mortality reductions between screening and control arms. These
mortality reductions result from a combination of the screening regimen,
screening technology and the effect of the early, screening-induced, treatment.
This motivates addressing these different aspects separately. Here we are
interested in the causal effect of early versus delayed treatments on cancer
mortality among the screening-detectable subgroup, which under certain
assumptions is estimable from conventional randomized screening trial using
instrumental variable type methods. To define the causal effect of interest, we
formulate a simplified structural multi-state model for screening trials, based
on a hypothetical intervention trial where screening detected individuals would
be randomized into early versus delayed treatments. The cancer-specific
mortality reductions after screening detection are quantified by a
cause-specific hazard ratio. For this, we propose two estimators, based on an
estimating equation and a likelihood expression. The methods extend existing
instrumental variable methods for time-to-event and competing risks outcomes to
time-dependent intermediate variables. Using the multi-state model as the basis
of a data generating mechanism, we investigate the performance of the new
estimators through simulation studies. In addition, we illustrate the proposed
method in the context of CT screening for lung cancer using the US National
Lung Screening Trial (NLST) data.
| stat |
Adaptative significance levels in linear regression models with known
variance | The Full Bayesian Significance Test (FBST) for precise hypotheses was
presented by Pereira and Stern [Entropy 1(4) (1999) 99-110] as a Bayesian
alternative instead of the traditional significance test using p-value. The
FBST is based on the evidence in favor of the null hypothesis (H). An important
practical issue for the implementation of the FBST is the determination of how
large the evidence must be in order to decide for its rejection. In the
Classical significance tests, it is known that p-value decreases as sample size
increases, so by setting a single significance level, it usually leads H
rejection. In the FBST procedure, the evidence in favor of H exhibits the same
behavior as the p-value when the sample size increases. This suggests that the
cut-off point to define the rejection of H in the FBST should be a sample size
function. In this work, the scenario of Linear Regression Models with known
variance under the Bayesian approach is considered, and a method to find a
cut-off value for the evidence in the FBST is presented by minimizing the
linear combination of the averaged type I and type II error probabilities for a
given sample size and also for a given dimension of the parametric space.
| stat |
Optimal Sampling Regimes for Estimating Population Dynamics | Ecologists are interested in modeling the population growth of species in
various ecosystems. Studying population dynamics can assist environmental
managers in making better decisions for the environment. Traditionally, the
sampling of species and tracking of populations have been recorded on a regular
time frequency. However, sampling can be an expensive process due to available
resources, money and time. Limiting sampling makes it challenging to properly
track the growth of a population. Thus, we propose a new and novel approach to
designing sampling regimes based on the dynamics associated with population
growth models. This design study minimizes the amount of time ecologists spend
in the field, while maximizing the information provided by the data.
| stat |
On community structure validation in real networks | Community structure is a commonly observed feature of real networks. The term
refers to the presence in a network of groups of nodes (communities) that
feature high internal connectivity, but are poorly connected between each
other. Whereas the issue of community detection has been addressed in several
works, the problem of validating a partition of nodes as a good community
structure for a real network has received considerably less attention and
remains an open issue. We propose a set of indices for community structure
validation of network partitions, which rely on concepts from network
enrichment analysis. The proposed indices allow to compare the adequacy of
different partitions of nodes as community structures. Moreover, they can be
employed to assess whether two networks share the same or similar community
structures, and to evaluate the performance of different network clustering
algorithms.
| stat |
Connecting the Dots: Towards Continuous Time Hamiltonian Monte Carlo | Continuous time Hamiltonian Monte Carlo is introduced, as a powerful
alternative to Markov chain Monte Carlo methods for continuous target
distributions. The method is constructed in two steps: First Hamiltonian
dynamics are chosen as the deterministic dynamics in a continuous time
piecewise deterministic Markov process. Under very mild restrictions, such a
process will have the desired target distribution as an invariant distribution.
Secondly, the numerical implementation of such processes, based on adaptive
numerical integration of second order ordinary differential equations is
considered. The numerical implementation yields an approximate, yet highly
robust algorithm that, unlike conventional Hamiltonian Monte Carlo, enables the
exploitation of the complete Hamiltonian trajectories (hence the title). The
proposed algorithm may yield large speedups and improvements in stability
relative to relevant benchmarks, while incurring numerical errors that are
negligible relative to the overall Monte Carlo errors.
| stat |
Optimal Paired Comparison Block Designs | For paired comparison experiments involving competing options described by
two-level attributes several different methods of constructing designs having
block paired observations under the main effects model are presented. These
designs are compared to alternative designs available in the literature.
| stat |
Adaptive Monte Carlo Multiple Testing via Multi-Armed Bandits | Monte Carlo (MC) permutation test is considered the gold standard for
statistical hypothesis testing, especially when standard parametric assumptions
are not clear or likely to fail. However, in modern data science settings where
a large number of hypothesis tests need to be performed simultaneously, it is
rarely used due to its prohibitive computational cost. In genome-wide
association studies, for example, the number of hypothesis tests $m$ is around
$10^6$ while the number of MC samples $n$ for each test could be greater than
$10^8$, totaling more than $nm$=$10^{14}$ samples. In this paper, we propose
Adaptive MC multiple Testing (AMT) to estimate MC p-values and control false
discovery rate in multiple testing. The algorithm outputs the same result as
the standard full MC approach with high probability while requiring only
$\tilde{O}(\sqrt{n}m)$ samples. This sample complexity is shown to be optimal.
On a Parkinson GWAS dataset, the algorithm reduces the running time from 2
months for full MC to an hour. The AMT algorithm is derived based on the theory
of multi-armed bandits.
| stat |
Estimation and prediction of Gaussian processes using generalized Cauchy
covariance model under fixed domain asymptotics | We study estimation and prediction of Gaussian processes with covariance
model belonging to the generalized Cauchy (GC) family, under fixed domain
asymptotics. Gaussian processes with this kind of covariance function provide
separate characterization of fractal dimension and long range dependence, an
appealing feature in many physical, biological or geological systems. The
results of the paper are classified into three parts. In the first part, we
characterize the equivalence of two Gaussian measures with GC covariance
function. Then we provide sufficient conditions for the equivalence of two
Gaussian measures with Mat{\'e}rn (MT) and GC covariance functions and two
Gaussian measures with Generalized Wendland (GW) and GC covariance functions.
In the second part, we establish strong consistency and asymptotic distribution
of the maximum likelihood estimator of the microergodic parameter associated to
GC covariance model, under fixed domain asymptotics. The last part focuses on
optimal prediction with GC model and specifically, we give conditions for
asymptotic efficiency prediction and asymptotically correct estimation of mean
square error using a misspecified GC, MT or GW model, under fixed domain
asymptotics. Our findings are illustrated through a simulation study: the first
compares the finite sample behavior of the maximum likelihood estimation of the
microergodic parameter of the GC model with the given asymptotic distribution.
We then compare the finite-sample behavior of the prediction and its associated
mean square error when the true model is GC and the prediction is performed
using the true model and a misspecified GW model.
| stat |
Covariate-dependent control limits for the detection of abnormal price
changes in scanner data | Currently, large-scale sales data for consumer goods, called scanner data,
are obtained by scanning the bar codes of individual products at the points of
sale of retail outlets. Many national statistical offices use scanner data to
build consumer price statistics. In this process, as in other statistical
procedures, the detection of abnormal transactions in sales prices is an
important step in the analysis. Popular methods for conducting such outlier
detection are the quartile method, the Hidiroglou-Berthelot method, the
resistant fences method, and the Tukey algorithm. These methods are based
solely on information about price changes and not on any of the other
covariates (e.g., sales volume or types of retail shops) that are also
available from scanner data. In this paper, we propose a new method to detect
abnormal price changes that takes into account an additional covariate, namely,
sales volume. We assume that the variance of the log of the price change is a
smooth function of the sales volume and estimate the function from previously
observed data. We numerically show the advantages of the new method over
existing methods. We also apply the methods to real scanner data collected at
weekly intervals by the Korean Chamber of Commerce and Industry between 2013
and 2014 and compare their performance.
| stat |
Bayesian Variable Selection for Multi-Outcome Models Through Shared
Shrinkage | Variable selection over a potentially large set of covariates in a linear
model is quite popular. In the Bayesian context, common prior choices can lead
to a posterior expectation of the regression coefficients that is a sparse (or
nearly sparse) vector with a few non-zero components, those covariates that are
most important. This article extends the global-local shrinkage idea to a
scenario where one wishes to model multiple response variables simultaneously.
Here, we have developed a variable selection method for a K-outcome model
(multivariate regression) that identifies the most important covariates across
all outcomes. The prior for all regression coefficients is a mean zero normal
with coefficient-specific variance term that consists of a predictor-specific
factor (shared local shrinkage parameter) and a model-specific factor (global
shrinkage term) that differs in each model. The performance of our modeling
approach is evaluated through simulation studies and a data example.
| stat |
Spatial Flow-Field Approximation Using Few Thermodynamic Measurements
Part II: Uncertainty Assessments | In this second part of our two-part paper, we provide a detailed, frequentist
framework for propagating uncertainties within our multivariate linear least
squares model. This permits us to quantify the impact of uncertainties in
thermodynamic measurements---arising from calibrations and the data acquisition
system---and the correlations therein, along with uncertainties in probe
positions. We show how the former has a much larger effect (relatively) than
uncertainties in probe placement.
We use this non-deterministic framework to demonstrate why the well-worn
metric for assessing spatial sampling uncertainty falls short of providing an
accurate characterization of the effect of a few spatial measurements. In other
words, it does not accurately describe the uncertainty associated with sampling
a non-uniform pattern with a few circumferentially scattered rakes. To this
end, we argue that our data-centric framework can offer a more rigorous
characterization of this uncertainty. Our paper proposes two new uncertainty
metrics: one for characterizing spatial sampling uncertainty and another for
capturing the impact of measurement imprecision in individual probes. These
metrics are rigorously derived in our paper and their ease in computation
permits them to be widely adopted by the turbomachinery community for carrying
out uncertainty assessments.
| stat |
A Smoothed P-Value Test When There is a Nuisance Parameter under the
Alternative | We present a new test when there is a nuisance parameter under the
alternative hypothesis. The test exploits the p-value occupation time [PVOT],
the measure of the subset of a nuisance parameter on which a p-value test
rejects the null hypothesis. Key contributions are: (i) An asymptotic critical
value upper bound for our test is the significance level, making inference
easy. Conversely, test statistic functionals need a bootstrap or simulation
step which can still lead to size and power distortions, and bootstrapped or
simulated critical values are not asymptotically valid under weak or
non-identification. (ii) We only require the test statistic to have a known or
bootstrappable limit distribution, hence we do not require root(n)-Gaussian
asymptotics, and weak or non-identification is allowed. Finally, (iii) a test
based on the sup-p-value may be conservative and in some cases have nearly
trivial power, while the PVOT naturally controls for this by smoothing over the
nuisance parameter space. We give examples and related controlled experiments
concerning PVOT tests of: omitted nonlinearity with and without the possibility
of weakly identified model parameters; GARCH effects. Across cases, the PVOT
test variously matches, dominates or strongly dominates standard tests based on
the supremum p-value, or supremum or average test statistic (with wild
bootstrapped p-value)
| stat |
Variational Inference to Measure Model Uncertainty in Deep Neural
Networks | We present a novel approach for training deep neural networks in a Bayesian
way. Classical, i.e. non-Bayesian, deep learning has two major drawbacks both
originating from the fact that network parameters are considered to be
deterministic. First, model uncertainty cannot be measured thus limiting the
use of deep learning in many fields of application and second, training of deep
neural networks is often hampered by overfitting. The proposed approach uses
variational inference to approximate the intractable a posteriori distribution
on basis of a normal prior. The variational density is designed in such a way
that the a posteriori uncertainty of the network parameters is represented per
network layer and depending on the estimated parameter expectation values. This
way, only a few additional parameters need to be optimized compared to a
non-Bayesian network. We apply this Bayesian approach to train and test the
LeNet architecture on the MNIST dataset. Compared to classical deep learning,
the test error is reduced by 15%. In addition, the trained model contains
information about the parameter uncertainty in each layer. We show that this
information can be used to calculate credible intervals for the prediction and
to optimize the network architecture for a given training data set.
| stat |
A surrogate model for estimating extreme tower loads on wind turbines
based on random forest proximities | In the present paper we present a surrogate model, which can be used to
estimate extreme tower loads on a wind turbine from a number of signals and a
suitable simulation tool. Due to the requirements of the International
Electrotechnical Commission (IEC) Standard 61400-1, assessing extreme tower
loads on wind turbines constitutes a key component of the design phase. The
proposed model imputes tower loads by matching observed signals with simulated
quantities using proximities induced by random forests. In this way the
algorithm's adaptability to high-dimensional and sparse settings is exploited
without using regression-based surrogate loads (which may display misleading
probabilistic characteristics). Finally, the model is applied to estimate tower
loads on an operating wind turbine from data on its operational statistics.
| stat |
A Joint Spatial Conditional Auto-Regressive Model for Estimating HIV
Prevalence Rates Among Key Populations | Ending the HIV/AIDS pandemic is among the Sustainable Development Goals for
the next decade. In order to overcome the gap between the need for care and the
available resources, better understanding of HIV epidemics is needed to guide
policy decisions, especially for key populations that are at higher risk for
HIV infection. Accurate HIV epidemic estimates for key populations have been
difficult to obtain because their HIV surveillance data is very limited. In
this paper, we propose a so-called joint spatial conditional auto-regressive
model for estimating HIV prevalence rates among key populations. Our model
borrows information from both neighboring locations and dependent populations.
As illustrated in the real data analysis, it provides more accurate estimates
than independently fitting the sub-epidemic for each key population. In
addition, we provide a study to reveal the conditions that our proposal gives a
better prediction. The study combines both theoretical investigation and
numerical study, revealing strength and limitations of our proposal.
| stat |
Compositionally-Warped Gaussian Processes | The Gaussian process (GP) is a nonparametric prior distribution over
functions indexed by time, space, or other high-dimensional index set. The GP
is a flexible model yet its limitation is given by its very nature: it can only
model Gaussian marginal distributions. To model non-Gaussian data, a GP can be
warped by a nonlinear transformation (or warping) as performed by warped GPs
(WGPs) and more computationally-demanding alternatives such as Bayesian WGPs
and deep GPs. However, the WGP requires a numerical approximation of the
inverse warping for prediction, which increases the computational complexity in
practice. To sidestep this issue, we construct a novel class of warpings
consisting of compositions of multiple elementary functions, for which the
inverse is known explicitly. We then propose the compositionally-warped GP
(CWGP), a non-Gaussian generative model whose expressiveness follows from its
deep compositional architecture, and its computational efficiency is guaranteed
by the analytical inverse warping. Experimental validation using synthetic and
real-world datasets confirms that the proposed CWGP is robust to the choice of
warpings and provides more accurate point predictions, better trained models
and shorter computation times than WGP.
| stat |
Second-order Control of Complex Systems with Correlated Synthetic Data | The generation of synthetic data is an essential tool to study complex
systems, allowing for example to test models of these in precisely controlled
settings, or to parametrize simulation models when data is missing. This paper
focuses on the generation of synthetic data with an emphasis on correlation
structure. We introduce a new methodology to generate such correlated synthetic
data. It is implemented in the field of socio-spatial systems, more precisely
by coupling an urban growth model with a transportation network generation
model. We also show the genericity of the method with an application on
financial time-series. The simulation results show that the generation of
correlated synthetic data for such systems is indeed feasible within a broad
range of correlations, and suggest applications of such synthetic datasets.
| stat |
A Nested K-Nearest Prognostic Approach for Microwave Precipitation Phase
Detection over Snow Cover | Monitoring changes of precipitation phase from space is important for
understanding the mass balance of Earth's cryosphere in a changing climate.
This paper examines a Bayesian nearest neighbor approach for prognostic
detection of precipitation and its phase using passive microwave observations
from the Global Precipitation Measurement (GPM) satellite. The method uses the
weighted Euclidean distance metric to search through an a priori database
populated with coincident GPM radiometer and radar observations as well as
ancillary snow-cover data. The algorithm performance is evaluated using data
from GPM official precipitation products, ground-based radars, and
high-fidelity simulations from the Weather Research and Forecasting model.
Using the presented approach, we demonstrate that the hit probability of
terrestrial precipitation detection can reach to 0.80, while the probability of
false alarm remains below 0.11. The algorithm demonstrates higher skill in
detecting snowfall than rainfall, on average by 10 percent. In particular, the
probability of precipitation detection and its solid phase increases by 11 and
8 percent, over dry snow cover, when compared to other surface types. The main
reason is found to be related to the ability of the algorithm in capturing the
signal of increased liquid water content in snowy clouds over radiometrically
cold snow-covered surfaces
| stat |
A functional-model-adjusted spatial scan statistic | This paper introduces a new spatial scan statistic designed to adjust cluster
detection for longitudinal confounding factors indexed in space. The
functional-model-adjusted statistic was developed using generalized functional
linear models in which longitudinal confounding factors were considered to be
functional covariates. A general framework was developed for application to
various probability models. Application to a Poisson model showed that the new
method is equivalent to a conventional spatial scan statistic that adjusts the
underlying population for covariates. In a simulation study with univariate and
multivariate models, we found that our new method adjusts the cluster detection
procedure more accurately than other methods. Use of the new spatial scan
statistic was illustrated by analysing data on premature mortality in France
over the period from 1998 to 2013, with the quarterly unemployment rate as a
longitudinal confounding factor.
| stat |
A defensive marginal particle filtering method for data assimilation | Particle filtering (PF) is an often used method to estimate the states of
dynamical systems. A major limitation of the standard PF method is that the
dimensionality of the state space increases as the time proceeds and eventually
may cause degeneracy of the algorithm. A possible approach to alleviate the
degeneracy issue is to compute the marginal posterior distribution at each time
step, which leads to the so-called marginal PF method. A key issue in the
marginal PF method is to construct a good sampling distribution in the marginal
space. When the posterior distribution is close to Gaussian, the Ensemble
Kalman filter (EnKF) method can usually provide a good sampling distribution;
however the EnKF approximation may fail completely when the posterior is
strongly non-Gaussian. In this work we propose a defensive marginal PF (DMPF)
algorithm which constructs a sampling distribution in the marginal space by
combining the standard PF and the EnKF approximation using a multiple
importance sampling (MIS) scheme. An important feature of the proposed
algorithm is that it can automatically adjust the relative weight of the PF and
the EnKF components in the MIS scheme in each step, according to how
non-Gaussian the posterior is. With numerical examples we demonstrate that the
proposed method can perform well regardless of whether the posteriors can be
well approximated by Gaussian.
| stat |
Estimation of the Number of Components of Non-Parametric Multivariate
Finite Mixture Models | We propose a novel estimator for the number of components (denoted by $M$) in
a K-variate non-parametric finite mixture model, where the analyst has repeated
observations of $K\geq2$ variables that are independent given a finitely
supported unobserved variable. Under a mild assumption on the joint
distribution of the observed and latent variables, we show that an integral
operator $T$, that is identified from the data, has rank equal to $M$. Using
this observation, and the fact that singular values are stable under
perturbations, the estimator of $M$ that we propose is based on a thresholding
rule which essentially counts the number of singular values of a consistent
estimator of $T$ that are greater than a data-driven threshold. We prove that
our estimator of $M$ is consistent, and establish non-asymptotic results which
provide finite sample performance guarantees for our estimator. We present a
Monte Carlo study which shows that our estimator performs well for samples of
moderate size.
| stat |
An Efficient Algorithm for Capacity-Approaching Noisy Adaptive Group
Testing | In this paper, we consider the group testing problem with adaptive test
designs and noisy outcomes. We propose a computationally efficient four-stage
procedure with components including random binning, identification of bins
containing defective items, 1-sparse recovery via channel codes, and a
"clean-up" step to correct any errors from the earlier stages. We prove that
the asymptotic required number of tests comes very close to the best known
information-theoretic achievability bound (which is based on computationally
intractable decoding), and approaches a capacity-based converse bound in the
low-sparsity regime.
| stat |
No Bad Local Minima in Wide and Deep Nonlinear Neural Networks | In this paper, we prove that no bad local minimum exists in the deep
nonlinear neural networks with sufficiently large widths of the hidden layers
if the parameters are initialized by He initialization method. Specifically, in
the deep ReLU neural network model with sufficiently large widths of the hidden
layers, the following four statements hold true: 1) the loss function is
non-convex and non-concave; 2) every local minimum is a global minimum; 3)
every critical point that is not a global minimum is a saddle point; and 4) bad
saddle points exist.
| stat |
Direct Estimation of Differential Functional Graphical Models | We consider the problem of estimating the difference between two functional
undirected graphical models with shared structures. In many applications, data
are naturally regarded as high-dimensional random function vectors rather than
multivariate scalars. For example, electroencephalography (EEG) data are more
appropriately treated as functions of time. In these problems, not only can the
number of functions measured per sample be large, but each function is itself
an infinite dimensional object, making estimation of model parameters
challenging. We develop a method that directly estimates the difference of
graphs, avoiding separate estimation of each graph, and show it is consistent
in certain high-dimensional settings. We illustrate finite sample properties of
our method through simulation studies. Finally, we apply our method to EEG data
to uncover differences in functional brain connectivity between alcoholics and
control subjects.
| stat |
Churn Prediction with Sequential Data and Deep Neural Networks. A
Comparative Analysis | Off-the-shelf machine learning algorithms for prediction such as regularized
logistic regression cannot exploit the information of time-varying features
without previously using an aggregation procedure of such sequential data.
However, recurrent neural networks provide an alternative approach by which
time-varying features can be readily used for modeling. This paper assesses the
performance of neural networks for churn modeling using recency, frequency, and
monetary value data from a financial services provider. Results show that RFM
variables in combination with LSTM neural networks have larger top-decile lift
and expected maximum profit metrics than regularized logistic regression models
with commonly-used demographic variables. Moreover, we show that using the
fitted probabilities from the LSTM as feature in the logistic regression
increases the out-of-sample performance of the latter by 25 percent compared to
a model with only static features.
| stat |
Second-order variational equations for spatial point processes with a
view to pair correlation function estimation | Second-order variational type equations for spatial point processes are
established. In case of log linear parametric models for pair correlation
functions, it is demonstrated that the variational equations can be applied to
construct estimating equations with closed form solutions for the parameter
estimates. This result is used to fit orthogonal series expansions of log pair
correlation functions of general form.
| stat |
Item Response Theory based Ensemble in Machine Learning | In this article, we propose a novel probabilistic framework to improve the
accuracy of a weighted majority voting algorithm. In order to assign higher
weights to the classifiers which can correctly classify hard-to-classify
instances, we introduce the Item Response Theory (IRT) framework to evaluate
the samples' difficulty and classifiers' ability simultaneously. Three models
are created with different assumptions suitable for different cases. When
making an inference, we keep a balance between the accuracy and complexity. In
our experiment, all the base models are constructed by single trees via
bootstrap. To explain the models, we illustrate how the IRT ensemble model
constructs the classifying boundary. We also compare their performance with
other widely used methods and show that our model performs well on 19 datasets.
| stat |
An Exact Auxiliary Variable Gibbs Sampler for a Class of Diffusions | Stochastic differential equations (SDEs) or diffusions are continuous-valued
continuous-time stochastic processes widely used in the applied and
mathematical sciences. Simulating paths from these processes is usually an
intractable problem, and typically involves time-discretization approximations.
We propose an exact Markov chain Monte Carlo sampling algorithm that involves
no such time-discretization error. Our sampler is applicable to the problem of
prior simulation from an SDE, posterior simulation conditioned on noisy
observations, as well as parameter inference given noisy observations. Our work
recasts an existing rejection sampling algorithm for a class of diffusions as a
latent variable model, and then derives an auxiliary variable Gibbs sampling
algorithm that targets the associated joint distribution. At a high level, the
resulting algorithm involves two steps: simulating a random grid of times from
an inhomogeneous Poisson process, and updating the SDE trajectory conditioned
on this grid. Our work allows the vast literature of Monte Carlo sampling
algorithms from the Gaussian process literature to be brought to bear to
applications involving diffusions. We study our method on synthetic and real
datasets, where we demonstrate superior performance over competing methods.
| stat |
Forecasting under model uncertainty:Non-homogeneous hidden Markov models
with Polya-Gamma data augmentation | We consider two-state Non-Homogeneous Hidden Markov Models (NHHMMs) for
forecasting univariate time series. Given a set of predictors, the time series
are modeled via predictive regressions with state dependent coefficients and
time-varying transition probabilities that depend on the predictors via a
logistic function. In a hidden Markov setting, inference for logistic
regression coefficients becomes complicated and in some cases impossible due to
convergence issues. In this paper, we aim to address this problem using a new
latent variable scheme that utilizes the P\'{o}lya-Gamma class of
distributions. We allow for model uncertainty regarding the predictors that
affect the series both linearly -- in the mean -- and non-linearly -- in the
transition matrix. Predictor selection and inference on the model parameters
are based on a MCMC scheme with reversible jump steps. Single-step and
multiple-steps-ahead predictions are obtained by the most probable model,
median probability model or a Bayesian Model Averaging approach. Using
simulation experiments, we illustrate the performance of our algorithm in
various setups, in terms of mixing properties, model selection and predictive
ability. An empirical study on realized volatility data shows that our
methodology gives improved forecasts compared to benchmark models.
| stat |
Manifold Regularization for Locally Stable Deep Neural Networks | We apply concepts from manifold regularization to develop new regularization
techniques for training locally stable deep neural networks. Our regularizers
are based on a sparsification of the graph Laplacian which holds with high
probability when the data is sparse in high dimensions, as is common in deep
learning. Empirically, our networks exhibit stability in a diverse set of
perturbation models, including $\ell_2$, $\ell_\infty$, and Wasserstein-based
perturbations; in particular, we achieve 40% adversarial accuracy on CIFAR-10
against an adaptive PGD attack using $\ell_\infty$ perturbations of size
$\epsilon = 8/255$, and state-of-the-art verified accuracy of 21% in the same
perturbation model. Furthermore, our techniques are efficient, incurring
overhead on par with two additional parallel forward passes through the
network.
| stat |
Joint spatial modeling of significant wave height and wave period using
the SPDE approach | The ocean wave distribution in a specific region of space and time is
described by its sea state. Knowledge about the sea states a ship encounters on
a journey can be used to assess various parameters of risk and wear associated
with the journey.
Two important characteristics of the sea state are the significant wave
height and mean wave period. We propose a joint spatial model of these two
quantities on the north Atlantic ocean. The model describes the distribution of
the logarithm of the two quantities as a bivariate Gaussian random field. This
random field is modeled as a solution to a system of coupled stochastic partial
differential equations. The bivariate random field can model a wide variety of
non-stationary anisotropy and allows for arbitrary, and different,
differentiability for the two marginal fields.
The parameters of the model are estimated on data of the north Atlantic using
a stepwise maximum likelihood method. The fitted model is used to derive the
distribution of accumulated fatigue damage for a ship sailing a transatlantic
route. Also, a method for estimating the risk of capsizing due to broaching-to,
based on the joint distribution of the two sea state characteristics, is
investigated. The risks are calculated for a transatlantic route between
America and Europe using both data and the fitted model.
The results show that the model compares well with observed data. Also, it
shows that the bivariate model is needed and cannot simply be approximated by a
model of significant wave height alone.
| stat |
DoWhy: An End-to-End Library for Causal Inference | In addition to efficient statistical estimators of a treatment's effect,
successful application of causal inference requires specifying assumptions
about the mechanisms underlying observed data and testing whether they are
valid, and to what extent. However, most libraries for causal inference focus
only on the task of providing powerful statistical estimators. We describe
DoWhy, an open-source Python library that is built with causal assumptions as
its first-class citizens, based on the formal framework of causal graphs to
specify and test causal assumptions. DoWhy presents an API for the four steps
common to any causal analysis---1) modeling the data using a causal graph and
structural assumptions, 2) identifying whether the desired effect is estimable
under the causal model, 3) estimating the effect using statistical estimators,
and finally 4) refuting the obtained estimate through robustness checks and
sensitivity analyses. In particular, DoWhy implements a number of robustness
checks including placebo tests, bootstrap tests, and tests for unoberved
confounding. DoWhy is an extensible library that supports interoperability with
other implementations, such as EconML and CausalML for the the estimation step.
The library is available at https://github.com/microsoft/dowhy
| stat |
Combining Prediction Intervals on Multi-Source Non-Disclosed Regression
Datasets | Conformal Prediction is a framework that produces prediction intervals based
on the output from a machine learning algorithm. In this paper we explore the
case when training data is made up of multiple parts available in different
sources that cannot be pooled. We here consider the regression case and propose
a method where a conformal predictor is trained on each data source
independently, and where the prediction intervals are then combined into a
single interval. We call the approach Non-Disclosed Conformal Prediction
(NDCP), and we evaluate it on a regression dataset from the UCI machine
learning repository using support vector regression as the underlying machine
learning algorithm, with varying number of data sources and sizes. The results
show that the proposed method produces conservatively valid prediction
intervals, and while we cannot retain the same efficiency as when all data is
used, efficiency is improved through the proposed approach as compared to
predicting using a single arbitrarily chosen source.
| stat |
The frontier of simulation-based inference | Many domains of science have developed complex simulations to describe
phenomena of interest. While these simulations provide high-fidelity models,
they are poorly suited for inference and lead to challenging inverse problems.
We review the rapidly developing field of simulation-based inference and
identify the forces giving new momentum to the field. Finally, we describe how
the frontier is expanding so that a broad audience can appreciate the profound
change these developments may have on science.
| stat |
Leveraging Semantic Embeddings for Safety-Critical Applications | Semantic Embeddings are a popular way to represent knowledge in the field of
zero-shot learning. We observe their interpretability and discuss their
potential utility in a safety-critical context. Concretely, we propose to use
them to add introspection and error detection capabilities to neural network
classifiers. First, we show how to create embeddings from symbolic domain
knowledge. We discuss how to use them for interpreting mispredictions and
propose a simple error detection scheme. We then introduce the concept of
semantic distance: a real-valued score that measures confidence in the semantic
space. We evaluate this score on a traffic sign classifier and find that it
achieves near state-of-the-art performance, while being significantly faster to
compute than other confidence scores. Our approach requires no changes to the
original network and is thus applicable to any task for which domain knowledge
is available.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.