title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Identification Risks Evaluation of Partially Synthetic Data with the
$\texttt{IdentificationRiskCalculation}$ R Package | We extend a general approach to evaluating identification risk of synthesized
variables in partially synthetic data. For multiple continuous synthesized
variables, we introduce the use of a radius $r$ in the construction of
identification risk probability of each target record, and illustrate with
working examples. We create the $\texttt{IdentificationRiskCalculation}$ R
package to aid researchers and data disseminators in performing these
identification risks evaluation calculations. We demonstrate our methods
through the R package with applications to a data sample from the Consumer
Expenditure Surveys, and discuss the impacts on risk and data utility of 1) the
choice of radius $r$, 2) the choice of synthesized variables, and 3) the choice
of number of synthetic datasets. We give recommendations for statistical
agencies for synthesizing and evaluating identification risk of continuous
variables.
| stat |
ARSM: Augment-REINFORCE-Swap-Merge Estimator for Gradient
Backpropagation Through Categorical Variables | To address the challenge of backpropagating the gradient through categorical
variables, we propose the augment-REINFORCE-swap-merge (ARSM) gradient
estimator that is unbiased and has low variance. ARSM first uses variable
augmentation, REINFORCE, and Rao-Blackwellization to re-express the gradient as
an expectation under the Dirichlet distribution, then uses variable swapping to
construct differently expressed but equivalent expectations, and finally shares
common random numbers between these expectations to achieve significant
variance reduction. Experimental results show ARSM closely resembles the
performance of the true gradient for optimization in univariate settings;
outperforms existing estimators by a large margin when applied to categorical
variational auto-encoders; and provides a "try-and-see self-critic" variance
reduction method for discrete-action policy gradient, which removes the need of
estimating baselines by generating a random number of pseudo actions and
estimating their action-value functions.
| stat |
Solving Bernoulli Rank-One Bandits with Unimodal Thompson Sampling | Stochastic Rank-One Bandits (Katarya et al, (2017a,b)) are a simple framework
for regret minimization problems over rank-one matrices of arms. The initially
proposed algorithms are proved to have logarithmic regret, but do not match the
existing lower bound for this problem. We close this gap by first proving that
rank-one bandits are a particular instance of unimodal bandits, and then
providing a new analysis of Unimodal Thompson Sampling (UTS), initially
proposed by Paladino et al (2017). We prove an asymptotically optimal regret
bound on the frequentist regret of UTS and we support our claims with
simulations showing the significant improvement of our method compared to the
state-of-the-art.
| stat |
Spectral Bias and Task-Model Alignment Explain Generalization in Kernel
Regression and Infinitely Wide Neural Networks | Generalization beyond a training dataset is a main goal of machine learning,
but theoretical understanding of generalization remains an open problem for
many models. The need for a new theory is exacerbated by recent observations in
deep neural networks where overparameterization leads to better performance,
contradicting the conventional wisdom from classical statistics. In this paper,
we investigate generalization error for kernel regression, which, besides being
a popular machine learning method, also includes infinitely overparameterized
neural networks trained with gradient descent. We use techniques from
statistical mechanics to derive an analytical expression for generalization
error applicable to any kernel or data distribution. We present applications of
our theory to real and synthetic datasets, and for many kernels including those
that arise from training deep neural networks in the infinite-width limit. We
elucidate an inductive bias of kernel regression to explain data with "simple
functions", which are identified by solving a kernel eigenfunction problem on
the data distribution. This notion of simplicity allows us to characterize
whether a kernel is compatible with a learning task, facilitating good
generalization performance from a small number of training examples. We show
that more data may impair generalization when noisy or not expressible by the
kernel, leading to non-monotonic learning curves with possibly many peaks. To
further understand these phenomena, we turn to the broad class of rotation
invariant kernels, which is relevant to training deep neural networks in the
infinite-width limit, and present a detailed mathematical analysis of them when
data is drawn from a spherically symmetric distribution and the number of input
dimensions is large.
| stat |
Convergence dynamics of Generative Adversarial Networks: the dual metric
flows | Fitting neural networks often resorts to stochastic (or similar) gradient
descent which is a noise-tolerant (and efficient) resolution of a gradient
descent dynamics. It outputs a sequence of networks parameters, which sequence
evolves during the training steps. The gradient descent is the limit, when the
learning rate is small and the batch size is infinite, of this set of
increasingly optimal network parameters obtained during training. In this
contribution, we investigate instead the convergence in the Generative
Adversarial Networks used in machine learning. We study the limit of small
learning rate, and show that, similar to single network training, the GAN
learning dynamics tend, for vanishing learning rate to some limit dynamics.
This leads us to consider evolution equations in metric spaces (which is the
natural framework for evolving probability laws) that we call dual flows. We
give formal definitions of solutions and prove the convergence. The theory is
then applied to specific instances of GANs and we discuss how this insight
helps understand and mitigate the mode collapse.
Keywords: GAN; metric flow; generative network
| stat |
Streamlined Variational Inference for Higher Level Group-Specific Curve
Models | A two-level group-specific curve model is such that the mean response of each
member of a group is a separate smooth function of a predictor of interest. The
three-level extension is such that one grouping variable is nested within
another one, and higher level extensions are analogous. Streamlined variational
inference for higher level group-specific curve models is a challenging
problem. We confront it by systematically working through two-level and then
three-level cases and making use of the higher level sparse matrix
infrastructure laid down in Nolan and Wand (2018). A motivation is analysis of
data from ultrasound technology for which three-level group-specific curve
models are appropriate. Whilst extension to the number of levels exceeding
three is not covered explicitly, the pattern established by our systematic
approach sheds light on what is required for even higher level group-specific
curve models.
| stat |
A Compressive Classification Framework for High-Dimensional Data | We propose a compressive classification framework for settings where the data
dimensionality is significantly higher than the sample size. The proposed
method, referred to as compressive regularized discriminant analysis (CRDA) is
based on linear discriminant analysis and has the ability to select significant
features by using joint-sparsity promoting hard thresholding in the
discriminant rule. Since the number of features is larger than the sample size,
the method also uses state-of-the-art regularized sample covariance matrix
estimators. Several analysis examples on real data sets, including image,
speech signal and gene expression data illustrate the promising improvements
offered by the proposed CRDA classifier in practise. Overall, the proposed
method gives fewer misclassification errors than its competitors, while at the
same time achieving accurate feature selection results. The open-source R
package and MATLAB toolbox of the proposed method (named compressiveRDA) is
freely available.
| stat |
Unbiased Smoothing using Particle Independent Metropolis-Hastings | We consider the approximation of expectations with respect to the
distribution of a latent Markov process given noisy measurements. This is known
as the smoothing problem and is often approached with particle and Markov chain
Monte Carlo (MCMC) methods. These methods provide consistent but biased
estimators when run for a finite time. We propose a simple way of coupling two
MCMC chains built using Particle Independent Metropolis-Hastings (PIMH) to
produce unbiased smoothing estimators. Unbiased estimators are appealing in the
context of parallel computing, and facilitate the construction of confidence
intervals. The proposed scheme only requires access to off-the-shelf Particle
Filters (PF) and is thus easier to implement than recently proposed unbiased
smoothers. The approach is demonstrated on a L\'evy-driven stochastic
volatility model and a stochastic kinetic model.
| stat |
Improved Grey System Models for Predicting Traffic Parameters | In transportation applications such as real-time route guidance, ramp
metering, congestion pricing and special events traffic management, accurate
short-term traffic flow prediction is needed. For this purpose, this paper
proposes several novel \textit{online} Grey system models (GM):
GM(1,1$|cos(\omega t)$), GM(1,1$|sin(\omega t)$, $cos(\omega t)$), and
GM(1,1$|e^{-at}$,$sin(\omega t)$,$cos(\omega t)$). To evaluate the performance
of the proposed models, they are compared against a set of benchmark models:
GM(1,1) model, Grey Verhulst models with and without Fourier error corrections,
linear time series model, and nonlinear time series model. The evaluation is
performed using loop detector and probe vehicle data from California, Virginia,
and Oregon. Among the benchmark models, the error corrected Grey Verhulst model
with Fourier outperformed the GM(1,1) model, linear time series, and non-linear
time series models. In turn, the three proposed models, GM(1,1$|cos(\omega
t)$), GM(1,1$|sin(\omega t)$,$cos(\omega t)$), and GM(1,1$|e^{-at}$,$sin(\omega
t)$,$cos(\omega t)$), outperformed the Grey Verhulst model in prediction by at
least $65\%$, $16\%$, and $11\%$, in terms of Root Mean Squared Error, and by
$82\%$, $58\%$, and $42\%$, in terms of Mean Absolute Percentage Error,
respectively. It is observed that the proposed Grey system models are more
adaptive to location (e.g., perform well for all roadway types) and traffic
parameters (e.g., speed, travel time, occupancy, and volume), and they do not
require as many data points for training (4 observations are found to be
sufficient).
| stat |
Bayesian model inversion using stochastic spectral embedding | In this paper we propose a new sampling-free approach to solve Bayesian model
inversion problems that is an extension of the previously proposed spectral
likelihood expansions (SLE) method. Our approach, called stochastic spectral
likelihood embedding (SSLE), uses the recently presented stochastic spectral
embedding (SSE) method for local spectral expansion refinement to approximate
the likelihood function at the core of Bayesian inversion problems. We show
that, similar to SLE, this approach results in analytical expressions for key
statistics of the Bayesian posterior distribution, such as evidence, posterior
moments and posterior marginals, by direct post-processing of the expansion
coefficients. Because SSLE and SSE rely on the direct approximation of the
likelihood function, they are in a way independent of the
computational/mathematical complexity of the forward model. We further enhance
the efficiency of SSLE by introducing a likelihood specific adaptive sample
enrichment scheme. To showcase the performance of the proposed SSLE, we solve
three problems that exhibit different kinds of complexity in the likelihood
function: multimodality, high posterior concentration and high nominal
dimensionality. We demonstrate how SSLE significantly improves on SLE, and
present it as a promising alternative to existing inversion frameworks.
| stat |
A comparison of apartment rent price prediction using a large dataset:
Kriging versus DNN | The hedonic approach based on a regression model has been widely adopted for
the prediction of real estate property price and rent. In particular, a spatial
regression technique called Kriging, a method of interpolation that was
advanced in the field of spatial statistics, are known to enable high accuracy
prediction in light of the spatial dependence of real estate property data.
Meanwhile, there has been a rapid increase in machine learning-based prediction
using a large (big) dataset and its effectiveness has been demonstrated in
previous studies. However, no studies have ever shown the extent to which
predictive accuracy differs for Kriging and machine learning techniques using
big data. Thus, this study compares the predictive accuracy of apartment rent
price in Japan between the nearest neighbor Gaussian processes (NNGP) model,
which enables application of Kriging to big data, and the deep neural network
(DNN), a representative machine learning technique, with a particular focus on
the data sample size (n = 10^4, 10^5, 10^6) and differences in predictive
performance. Our analysis showed that, with an increase in sample size, the
out-of-sample predictive accuracy of DNN approached that of NNGP and they were
nearly equal on the order of n = 10^6. Furthermore, it is suggested that, for
both higher and lower end properties whose rent price deviates from the median,
DNN may have a higher predictive accuracy than that of NNGP.
| stat |
A Global Stochastic Optimization Particle Filter Algorithm | We introduce a new algorithm to learn on the fly the parameter value
$\theta_\star:=\mathrm{argmax}_{\theta\in\Theta}\mathbb{E}[\log f_\theta(Y_0)]$
from a sequence $(Y_t)_{t\geq 1}$ of independent copies of $Y_0$, with
$\{f_\theta,\,\theta\in\Theta\subseteq\mathbb{R}^d\}$ a parametric model. The
main idea of the proposed approach is to define a sequence
$(\tilde{\pi}_t)_{t\geq 1}$ of probability distributions on $\Theta$ which (i)
is shown to concentrate on $\theta_\star$ as $t\rightarrow\infty$ and (ii) can
be estimated in an online fashion by means of a standard particle filter (PF)
algorithm. The sequence $(\tilde{\pi}_t)_{t\geq 1}$ depends on a learning rate
$h_t\rightarrow 0$, with the slower $h_t$ converges to zero the greater is the
ability of the PF approximation $\tilde{\pi}_t^N$ of $\tilde{\pi}_t$ to escape
from a local optimum of the objective function, but the slower is the rate at
which $\tilde{\pi}_t$ concentrates on $\theta_\star$. To conciliate ability to
escape from a local optimum and fast convergence towards $\theta_\star$ we
exploit the acceleration property of averaging, well-known in the stochastic
gradient descent literature, by letting $\bar{\theta}_t^N:=t^{-1}\sum_{s=1}^t
\int_{\Theta}\theta\ \tilde{\pi}_s^N(\mathrm{d} \theta)$ be the proposed
estimator of $\theta_\star$. Our numerical experiments suggest that
$\bar{\theta}_t^N$ converges to $\theta_\star$ at the optimal $t^{-1/2}$ rate
in challenging models and in situations where $\tilde{\pi}_t^N$ concentrates on
this parameter value at a slower rate. We illustrate the practical usefulness
of the proposed optimization algorithm for online parameter learning and for
computing the maximum likelihood estimator.
| stat |
The Augmented Synthetic Control Method | The synthetic control method (SCM) is a popular approach for estimating the
impact of a treatment on a single unit in panel data settings. The "synthetic
control" is a weighted average of control units that balances the treated
unit's pre-treatment outcomes as closely as possible. A critical feature of the
original proposal is to use SCM only when the fit on pre-treatment outcomes is
excellent. We propose Augmented SCM as an extension of SCM to settings where
such pre-treatment fit is infeasible. Analogous to bias correction for inexact
matching, Augmented SCM uses an outcome model to estimate the bias due to
imperfect pre-treatment fit and then de-biases the original SCM estimate. Our
main proposal, which uses ridge regression as the outcome model, directly
controls pre-treatment fit while minimizing extrapolation from the convex hull.
This estimator can also be expressed as a solution to a modified synthetic
controls problem that allows negative weights on some donor units. We bound the
estimation error of this approach under different data generating processes,
including a linear factor model, and show how regularization helps to avoid
over-fitting to noise. We demonstrate gains from Augmented SCM with extensive
simulation studies and apply this framework to estimate the impact of the 2012
Kansas tax cuts on economic growth. We implement the proposed method in the new
augsynth R package.
| stat |
A Discrete Bouncy Particle Sampler | Most Markov chain Monte Carlo methods operate in discrete time and are
reversible with respect to the target probability. Nevertheless, it is now
understood that the use of non-reversible Markov chains can be beneficial in
many contexts. In particular, the recently-proposed Bouncy Particle Sampler
leverages a continuous-time and non-reversible Markov process and empirically
shows state-of-the-art performances when used to explore certain probability
densities; however, its implementation typically requires the computation of
local upper bounds on the gradient of the log target density.
We present the Discrete Bouncy Particle Sampler, a general algorithm based
upon a guided random walk, a partial refreshment of direction, and a
delayed-rejection step. We show that the Bouncy Particle Sampler can be
understood as a scaling limit of a special case of our algorithm. In contrast
to the Bouncy Particle Sampler, implementing the Discrete Bouncy Particle
Sampler only requires point-wise evaluation of the target density and its
gradient. We propose extensions of the basic algorithm for situations when the
exact gradient of the target density is not available. In a Gaussian setting,
we establish a scaling limit for the radial process as dimension increases to
infinity. We leverage this result to obtain the theoretical efficiency of the
Discrete Bouncy Particle Sampler as a function of the partial-refreshment
parameter, which leads to a simple and robust tuning criterion. A further
analysis in a more general setting suggests that this tuning criterion applies
more generally. Theoretical and empirical efficiency curves are then compared
for different targets and algorithm variations.
| stat |
Task Agnostic Continual Learning via Meta Learning | While neural networks are powerful function approximators, they suffer from
catastrophic forgetting when the data distribution is not stationary. One
particular formalism that studies learning under non-stationary distribution is
provided by continual learning, where the non-stationarity is imposed by a
sequence of distinct tasks. Most methods in this space assume, however, the
knowledge of task boundaries, and focus on alleviating catastrophic forgetting.
In this work, we depart from this view and move the focus towards faster
remembering -- i.e measuring how quickly the network recovers performance
rather than measuring the network's performance without any adaptation. We
argue that in many settings this can be more effective and that it opens the
door to combining meta-learning and continual learning techniques, leveraging
their complementary advantages. We propose a framework specific for the
scenario where no information about task boundaries or task identity is given.
It relies on a separation of concerns into what task is being solved and how
the task should be solved. This framework is implemented by differentiating
task specific parameters from task agnostic parameters, where the latter are
optimized in a continual meta learning fashion, without access to multiple
tasks at the same time. We showcase this framework in a supervised learning
scenario and discuss the implication of the proposed formalism.
| stat |
Exploiting Low Rank Covariance Structures for Computing High-Dimensional
Normal and Student-$t$ Probabilities | We present a preconditioned Monte Carlo method for computing high-dimensional
multivariate normal and Student-$t$ probabilities arising in spatial
statistics. The approach combines a tile-low-rank representation of covariance
matrices with a block-reordering scheme for efficient Quasi-Monte Carlo
simulation. The tile-low-rank representation decomposes the high-dimensional
problem into many diagonal-block-size problems and low-rank connections. The
block-reordering scheme reorders between and within the diagonal blocks to
reduce the impact of integration variables from right to left, thus improving
the Monte Carlo convergence rate. Simulations up to dimension $65{,}536$
suggest that the new method can improve the run time by an order of magnitude
compared with the non-reordered tile-low-rank Quasi-Monte Carlo method and two
orders of magnitude compared with the dense Quasi-Monte Carlo method. Our
method also forms a strong substitute for the approximate conditioning methods
as a more robust estimation with error guarantees. An application study to wind
stochastic generators is provided to illustrate that the new computational
method makes the maximum likelihood estimation feasible for high-dimensional
skew-normal random fields.
| stat |
Uncertainty Propagation in Deep Neural Network Using Active Subspace | The inputs of deep neural network (DNN) from real-world data usually come
with uncertainties. Yet, it is challenging to propagate the uncertainty in the
input features to the DNN predictions at a low computational cost. This work
employs a gradient-based subspace method and response surface technique to
accelerate the uncertainty propagation in DNN. Specifically, the active
subspace method is employed to identify the most important subspace in the
input features using the gradient of the DNN output to the inputs. Then the
response surface within that low-dimensional subspace can be efficiently built,
and the uncertainty of the prediction can be acquired by evaluating the
computationally cheap response surface instead of the DNN models. In addition,
the subspace can help explain the adversarial examples. The approach is
demonstrated in MNIST datasets with a convolutional neural network. Code is
available at: https://github.com/jiweiqi/nnsubspace.
| stat |
Sequential Testing of Multinomial Hypotheses with Applications to
Detecting Implementation Errors and Missing Data in Randomized Experiments | Simply randomized designs are one of the most common controlled experiments
used to study causal effects. Failure of the assignment mechanism, to provide
proper randomization of units across treatments, or the data collection
mechanism, when data is missing not at random, can render subsequent analysis
invalid if not properly identified. In this paper we demonstrate that such
practical implementation errors can often be identified, fortunately, through
consideration of the total unit counts resulting in each treatment group. Based
on this observation, we introduce a sequential hypothesis test constructed from
Bayesian multinomial-Dirichlet families for detecting practical implementation
errors in simply randomized experiments. By establishing a Martingale property
of the posterior odds under the null hypothesis, frequentist Type-I error is
controlled under both optional stopping and continuation via maximal
inequalities, preventing practitioners from potentially inflating false
positive probabilities through continuous monitoring. In contrast to other
statistical tests that are performed once all data collection is completed, the
proposed test is sequential - frequently rejecting the null during the process
of data collection itself, saving further units from entering an
improperly-executed experiment. We illustrate the utility of this test in the
context of online controlled experiments (OCEs), where the assignment is
automated through code and data collected through complex processing pipelines,
often in the presence of unintended bugs and logical errors. Confidence
sequences possessing desired sequential frequentist coverage probabilities are
provided and their connection to the Bayesian support interval is examined. The
differences between pure Bayesian and sequential frequentist testing procedures
are finally discussed through a conditional frequentist testing perspective.
| stat |
Interim recruitment prediction for multi-centre clinical trials | We introduce a general framework for monitoring, modelling, and predicting
the recruitment to multi-centre clinical trials. The work is motivated by
overly optimistic and narrow prediction intervals produced by existing
time-homogeneous recruitment models for multi-centre recruitment. We first
present two tests for detection of decay in recruitment rates, together with a
power study. We then introduce a model based on the inhomogeneous Poisson
process with monotonically decaying intensity, motivated by recruitment trends
observed in oncology trials. The general form of the model permits adaptation
to any parametric curve-shape. A general method for constructing sensible
parameter priors is provided and Bayesian model averaging is used for making
predictions which account for the uncertainty in both the parameters and the
model. The validity of the method and its robustness to misspecification are
tested using simulated datasets. The new methodology is then applied to
oncology trial data, where we make interim accrual predictions, comparing them
to those obtained by existing methods, and indicate where unexpected changes in
the accrual pattern occur.
| stat |
Occam Factor for Gaussian Models With Unknown Variance Structure | We discuss model selection to determine whether the variance-covariance
matrix of a multivariate Gaussian model with known mean should be considered to
be a constant diagonal, a non-constant diagonal, or an arbitrary positive
definite matrix. Of particular interest is the relationship between Bayesian
evidence and the flexibility penalty due to Priebe and Rougier. For the case of
an exponential family in canonical form equipped with a conjugate prior for the
canonical parameter, flexibility may be exactly decomposed into the usual BIC
likelihood penalty and a $O_p(1)$ term, the latter of which we explicitly
compute. We also investigate the asymptotics of Bayes factors for linearly
nested canonical exponential families equipped with conjugate priors; in
particular, we find the exact rates at which Bayes factors correctly diverge in
favor of the correct model: linearly and logarithmically in the number of
observations when the full and nested models are true, respectively. Such
theoretical considerations for the general case permit us to fully express the
asymptotic behavior of flexibility and Bayes factors for the
variance-covariance structure selection problem when we assume that the prior
for the model precision is a member of the gamma/Wishart family of
distributions or is uninformative. Simulations demonstrate evidence's immediate
and superior performance in model selection compared to approximate criteria
such as the BIC. We extend the framework to the multivariate Gaussian linear
model with three data-driven examples.
| stat |
Using reference models in variable selection | Variable selection, or more generally, model reduction is an important aspect
of the statistical workflow aiming to provide insights from data. In this
paper, we discuss and demonstrate the benefits of using a reference model in
variable selection. A reference model acts as a noise-filter on the target
variable by modeling its data generating mechanism. As a result, using the
reference model predictions in the model selection procedure reduces the
variability and improves stability leading to improved model selection
performance. Assuming that a Bayesian reference model describes the true
distribution of future data well, the theoretically preferred usage of the
reference model is to project its predictive distribution to a reduced model
leading to projection predictive variable selection approach. Alternatively,
reference models may also be used in an ad-hoc manner in combination with
common variable selection methods. In several numerical experiments, we
investigate the performance of the projective prediction approach as well as
alternative variable selection methods with and without reference models. Our
results indicate that the use of reference models generally translates into
better and more stable variable selection. Additionally, we demonstrate that
the projection predictive approach shows superior performance as compared to
alternative variable selection methods independently of whether or not they use
reference models.
| stat |
A DFA-based bivariate regression model for estimating the dependence of
PM2.5 among neighbouring cities | On the basis of detrended fluctuation analysis (DFA), we propose a new
bivariate linear regression model. This new model provides estimators of
multi-scale regression coefficients to measure the dependence between variables
and corresponding variables of interest with multi-scales. Numerical tests are
performed to illustrate that the proposed DFA-based regression estimators are
capable of accurately depicting the dependence between the variables of
interest and can be used to identify different dependence at different time
scales. We apply this model to analyze the PM2.5 series of three adjacent
cities (Beijing, Tianjin, and Baoding) in Northern China. The estimated
regression coefficients confirmed the dependence of PM2.5 among the three
cities and illustrated that each city has different influence on the others at
different seasons and at different time scales. Two statistics based on the
scale-dependent $t$-statistic and the partial detrended cross-correlation
coefficient are used to demonstrate the significance of the dependence. Three
new scale-dependent evaluation indices show that the new DFA-based bivariate
regression model can provide rich information on studied variables.
| stat |
Real-time forecasts and risk assessment of novel coronavirus (COVID-19)
cases: A data-driven analysis | The coronavirus disease 2019 (COVID-19) has become a public health emergency
of international concern affecting 201 countries and territories around the
globe. As of April 4, 2020, it has caused a pandemic outbreak with more than
11,16,643 confirmed infections and more than 59,170 reported deaths worldwide.
The main focus of this paper is two-fold: (a) generating short term (real-time)
forecasts of the future COVID-19 cases for multiple countries; (b) risk
assessment (in terms of case fatality rate) of the novel COVID-19 for some
profoundly affected countries by finding various important demographic
characteristics of the countries along with some disease characteristics. To
solve the first problem, we presented a hybrid approach based on autoregressive
integrated moving average model and Wavelet-based forecasting model that can
generate short-term (ten days ahead) forecasts of the number of daily confirmed
cases for Canada, France, India, South Korea, and the UK. The predictions of
the future outbreak for different countries will be useful for the effective
allocation of health care resources and will act as an early-warning system for
government policymakers. In the second problem, we applied an optimal
regression tree algorithm to find essential causal variables that significantly
affect the case fatality rates for different countries. This data-driven
analysis will necessarily provide deep insights into the study of early risk
assessments for 50 immensely affected countries.
| stat |
Statistical Consequences of Fat Tails: Real World Preasymptotics,
Epistemology, and Applications | The monograph investigates the misapplication of conventional statistical
techniques to fat tailed distributions and looks for remedies, when possible.
Switching from thin tailed to fat tailed distributions requires more than
"changing the color of the dress". Traditional asymptotics deal mainly with
either n=1 or $n=\infty$, and the real world is in between, under of the "laws
of the medium numbers" --which vary widely across specific distributions. Both
the law of large numbers and the generalized central limit mechanisms operate
in highly idiosyncratic ways outside the standard Gaussian or Levy-Stable
basins of convergence.
A few examples:
+ The sample mean is rarely in line with the population mean, with effect on
"naive empiricism", but can be sometimes be estimated via parametric methods.
+ The "empirical distribution" is rarely empirical.
+ Parameter uncertainty has compounding effects on statistical metrics.
+ Dimension reduction (principal components) fails.
+ Inequality estimators (GINI or quantile contributions) are not additive and
produce wrong results.
+ Many "biases" found in psychology become entirely rational under more
sophisticated probability distributions
+ Most of the failures of financial economics, econometrics, and behavioral
economics can be attributed to using the wrong distributions.
This book, the first volume of the Technical Incerto, weaves a narrative
around published journal articles.
| stat |
Ensemble Methods to Assess Heterogeneity in Joint Development with
Nonlinear Trajectories | Researchers are often interested in uncovering heterogeneity in change
patterns and grouping trajectories into homogeneous latent classes. A
considerable number of theoretical and empirical studies have focused on
investigating heterogeneity in change patterns of a univariate repeated
outcome. However, developmental processes rarely unfold in isolation;
therefore, empirical researchers often desire to examine two or more outcomes
over time, hoping to understand their joint development where these outcomes
and their change patterns are correlated. It is also of importance to examine
the impact that covariates have on the heterogeneity of joint development. This
study examines the heterogeneity in parallel nonlinear trajectories and
identifies baseline characteristics as predictors of latent classes. Our
simulation studies show that the proposed model can separate parallel change
patterns and provide unbiased point estimates with small standard errors and
confidence intervals with satisfactory coverage probabilities for the
parameters of interest. We illustrate how to apply the model to investigate the
heterogeneity of parallel nonlinear trajectories of joint development of
reading and mathematics ability from Grade K to 5. In this real-world example,
we demonstrate how to employ two methods, feature selection and feature
reduction, to address covariate space with a large-dimension and highly
correlated subsets in the structural equation modeling framework.
| stat |
Robust testing in generalized linear models by sign-flipping score
contributions | Generalized linear models are often misspecified due to overdispersion,
heteroscedasticity and ignored nuisance variables. Existing quasi-likelihood
methods for testing in misspecified models often do not provide satisfactory
type-I error rate control. We provide a novel semi-parametric test, based on
sign-flipping individual score contributions. The tested parameter is allowed
to be multi-dimensional and even high-dimensional. Our test is often robust
against the mentioned forms of misspecification and provides better type-I
error control than its competitors. When nuisance parameters are estimated, our
basic test becomes conservative. We show how to take nuisance estimation into
account to obtain an asymptotically exact test. Our proposed test is
asymptotically equivalent to its parametric counterpart.
| stat |
Rates of Convergence for Sparse Variational Gaussian Process Regression | Excellent variational approximations to Gaussian process posteriors have been
developed which avoid the $\mathcal{O}\left(N^3\right)$ scaling with dataset
size $N$. They reduce the computational cost to $\mathcal{O}\left(NM^2\right)$,
with $M\ll N$ being the number of inducing variables, which summarise the
process. While the computational cost seems to be linear in $N$, the true
complexity of the algorithm depends on how $M$ must increase to ensure a
certain quality of approximation. We address this by characterising the
behavior of an upper bound on the KL divergence to the posterior. We show that
with high probability the KL divergence can be made arbitrarily small by
growing $M$ more slowly than $N$. A particular case of interest is that for
regression with normally distributed inputs in D-dimensions with the popular
Squared Exponential kernel, $M=\mathcal{O}(\log^D N)$ is sufficient. Our
results show that as datasets grow, Gaussian process posteriors can truly be
approximated cheaply, and provide a concrete rule for how to increase $M$ in
continual learning scenarios.
| stat |
Ensembled sparse-input hierarchical networks for high-dimensional
datasets | Neural networks have seen limited use in prediction for high-dimensional data
with small sample sizes, because they tend to overfit and require tuning many
more hyperparameters than existing off-the-shelf machine learning methods. With
small modifications to the network architecture and training procedure, we show
that dense neural networks can be a practical data analysis tool in these
settings. The proposed method, Ensemble by Averaging Sparse-Input Hierarchical
networks (EASIER-net), appropriately prunes the network structure by tuning
only two L1-penalty parameters, one that controls the input sparsity and
another that controls the number of hidden layers and nodes. The method selects
variables from the true support if the irrelevant covariates are only weakly
correlated with the response; otherwise, it exhibits a grouping effect, where
strongly correlated covariates are selected at similar rates. On a collection
of real-world datasets with different sizes, EASIER-net selected network
architectures in a data-adaptive manner and achieved higher prediction accuracy
than off-the-shelf methods on average.
| stat |
On the Stability and Generalization of Learning with Kernel Activation
Functions | In this brief we investigate the generalization properties of a
recently-proposed class of non-parametric activation functions, the kernel
activation functions (KAFs). KAFs introduce additional parameters in the
learning process in order to adapt nonlinearities individually on a per-neuron
basis, exploiting a cheap kernel expansion of every activation value. While
this increase in flexibility has been shown to provide significant improvements
in practice, a theoretical proof for its generalization capability has not been
addressed yet in the literature. Here, we leverage recent literature on the
stability properties of non-convex models trained via stochastic gradient
descent (SGD). By indirectly proving two key smoothness properties of the
models under consideration, we prove that neural networks endowed with KAFs
generalize well when trained with SGD for a finite number of steps.
Interestingly, our analysis provides a guideline for selecting one of the
hyper-parameters of the model, the bandwidth of the scalar Gaussian kernel. A
short experimental evaluation validates the proof.
| stat |
COVID-19 and other viruses: holding back its expansion by massive
testing | The experience of Singapur and South Korea makes it clear that under certain
circumstances massive testing is an effective way for containing the advance of
the COVID-19. In this paper, we propose a modified SEIR model which takes into
account tracing and massive testing, proving theoretically that more tracing
and testing implies a reduction of the total number of infected people in the
long run. We apply this model to the spread of the first wave of the disease in
Spain, obtaining numerical results. After that, we introduce a heuristic
approach in order to minimize the COVID-19 spreading by planning effective test
distributions among the populations of a region over a period of time. As an
application, the impact of distributing tests among the counties of New York
according to this method is computed in terms of the number of saved infected
individuals.
| stat |
A statistical theory of out-of-distribution detection | We introduce a principled approach to detecting out-of-distribution (OOD)
data by exploiting a connection to data curation. In data curation, we exclude
ambiguous or difficult-to-classify input points from the dataset, and these
excluded points are by definition OOD. We can therefore obtain the likelihood
for OOD points by using a principled generative model of data-curation
initially developed to explain the cold-posterior effect in Bayesian neural
networks (Aitchison 2020). This model gives higher OOD probabilities when
predictive uncertainty is higher and can be trained using maximum-likelihood
jointly over the in-distribution and OOD points. This approach gives superior
performance to past methods that did not provide a probability for OOD points,
and therefore could not be trained using maximum-likelihood.
| stat |
Temporal Parallelization of Bayesian Smoothers | This paper presents algorithms for temporal parallelization of Bayesian
smoothers. We define the elements and the operators to pose these problems as
the solutions to all-prefix-sums operations for which efficient parallel
scan-algorithms are available. We present the temporal parallelization of the
general Bayesian filtering and smoothing equations and specialize them to
linear/Gaussian models. The advantage of the proposed algorithms is that they
reduce the linear complexity of standard smoothing algorithms with respect to
time to logarithmic.
| stat |
Decomposable Probability-of-Success Metrics in Algorithmic Search | Previous studies have used a specific success metric within an algorithmic
search framework to prove machine learning impossibility results. However, this
specific success metric prevents us from applying these results on other forms
of machine learning, e.g. transfer learning. We define decomposable metrics as
a category of success metrics for search problems which can be expressed as a
linear operation on a probability distribution to solve this issue. Using an
arbitrary decomposable metric to measure the success of a search, we
demonstrate theorems which bound success in various ways, generalizing several
existing results in the literature.
| stat |
Learning Self-Imitating Diverse Policies | The success of popular algorithms for deep reinforcement learning, such as
policy-gradients and Q-learning, relies heavily on the availability of an
informative reward signal at each timestep of the sequential decision-making
process. When rewards are only sparsely available during an episode, or a
rewarding feedback is provided only after episode termination, these algorithms
perform sub-optimally due to the difficultly in credit assignment.
Alternatively, trajectory-based policy optimization methods, such as
cross-entropy method and evolution strategies, do not require per-timestep
rewards, but have been found to suffer from high sample complexity by
completing forgoing the temporal nature of the problem. Improving the
efficiency of RL algorithms in real-world problems with sparse or episodic
rewards is therefore a pressing need. In this work, we introduce a
self-imitation learning algorithm that exploits and explores well in the sparse
and episodic reward settings. We view each policy as a state-action visitation
distribution and formulate policy optimization as a divergence minimization
problem. We show that with Jensen-Shannon divergence, this divergence
minimization problem can be reduced into a policy-gradient algorithm with
shaped rewards learned from experience replays. Experimental results indicate
that our algorithm works comparable to existing algorithms in environments with
dense rewards, and significantly better in environments with sparse and
episodic rewards. We then discuss limitations of self-imitation learning, and
propose to solve them by using Stein variational policy gradient descent with
the Jensen-Shannon kernel to learn multiple diverse policies. We demonstrate
its effectiveness on a challenging variant of continuous-control MuJoCo
locomotion tasks.
| stat |
Adaptive Tuning Of Hamiltonian Monte Carlo Within Sequential Monte Carlo | Sequential Monte Carlo (SMC) samplers form an attractive alternative to MCMC
for Bayesian computation. However, their performance depends strongly on the
Markov kernels used to rejuvenate particles. We discuss how to calibrate
automatically (using the current particles) Hamiltonian Monte Carlo kernels
within SMC. To do so, we build upon the adaptive SMC approach of Fearnhead and
Taylor (2013), and we also suggest alternative methods. We illustrate the
advantages of using HMC kernels within an SMC sampler via an extensive
numerical study.
| stat |
Partial Optimal Transport with Applications on Positive-Unlabeled
Learning | Classical optimal transport problem seeks a transportation map that preserves
the total mass betwenn two probability distributions, requiring their mass to
be the same. This may be too restrictive in certain applications such as color
or shape matching, since the distributions may have arbitrary masses and/or
that only a fraction of the total mass has to be transported. Several
algorithms have been devised for computing partial Wasserstein metrics that
rely on an entropic regularization, but when it comes with exact solutions,
almost no partial formulation of neither Wasserstein nor Gromov-Wasserstein are
available yet. This precludes from working with distributions that do not lie
in the same metric space or when invariance to rotation or translation is
needed. In this paper, we address the partial Wasserstein and
Gromov-Wasserstein problems and propose exact algorithms to solve them. We
showcase the new formulation in a positive-unlabeled (PU) learning application.
To the best of our knowledge, this is the first application of optimal
transport in this context and we first highlight that partial Wasserstein-based
metrics prove effective in usual PU learning settings. We then demonstrate that
partial Gromov-Wasserstein metrics is efficient in scenario where point clouds
come from different domains or have different features.
| stat |
Scalable algorithms for semiparametric accelerated failure time models
in high dimensions | Semiparametric accelerated failure time (AFT) models are a useful alternative
to Cox proportional hazards models, especially when the assumption of constant
hazard ratios is untenable. However, rank-based criteria for fitting AFT models
are often non-differentiable, which poses a computational challenge in
high-dimensional settings. In this article, we propose a new alternating
direction method of multipliers algorithm for fitting semiparametric AFT models
by minimizing a penalized rank-based loss function. Our algorithm scales well
in both the number of subjects and number of predictors; and can easily
accommodate a wide range of popular penalties. To improve the selection of
tuning parameters, we propose a new criterion which avoids some common problems
in cross-validation with censored responses. Through extensive simulation
studies, we show that our algorithm and software is much faster than existing
methods (which can only be applied to special cases), and we show that
estimators which minimize a penalized rank-based criterion often outperform
alternative estimators which minimize penalized weighted least squares
criteria. Application to nine cancer datasets further demonstrates that
rank-based estimators of semiparametric AFT models are competitive with
estimators assuming proportional hazards model in high-dimensional settings,
whereas weighted least squares estimators are often not. A software package
implementing the algorithm, along with a set of auxiliary functions, is
available for download at github.com/ajmolstad/penAFT.
| stat |
Estimating finite mixtures of semi-Markov chains: an application to the
segmentation of temporal sensory data | In food science, it is of great interest to get information about the
temporal perception of aliments to create new products, to modify existing ones
or more generally to understand the perception mechanisms. Temporal Dominance
of Sensations (TDS) is a technique to measure temporal perception which
consists in choosing sequentially attributes describing a food product over
tasting. This work introduces new statistical models based on finite mixtures
of semi-Markov chains in order to describe data collected with the TDS
protocol, allowing different temporal perceptions for a same product within a
population. The identifiability of the parameters of such mixture models is
discussed. Sojourn time distributions are fitted with gamma probability
distribution and a penalty is added to the log likelihood to ensure convergence
of the EM algorithm to a non degenerate solution. Information criterions are
employed for determining the number of mixture components. Then, the individual
qualitative trajectories are clustered with the help of the maximum a
posteriori probability (MAP) approach. A simulation study confirms the good
behavior of the proposed estimation procedure. The methodology is illustrated
on an example of consumers perception of a Gouda cheese and assesses the
existence of several behaviors in terms of perception of this product.
| stat |
A Bayesian Approach to Block-Term Tensor Decomposition Model Selection
and Computation | The so-called block-term decomposition (BTD) tensor model, especially in its
rank-$(L_r,L_r,1)$ version, has been recently receiving increasing attention
due to its enhanced ability of representing systems and signals that are
composed of \emph{blocks} of rank higher than one, a scenario encountered in
numerous and diverse applications. Its uniqueness and approximation have thus
been thoroughly studied. Nevertheless, the challenging problem of estimating
the BTD model structure, namely the number of block terms and their individual
ranks, has only recently started to attract significant attention. In this
work, a Bayesian approach is taken to addressing the problem of
rank-$(L_r,L_r,1)$ BTD model selection and computation, based on the idea of
imposing column sparsity \emph{jointly} on the factors and in a
\emph{hierarchical} manner and estimating the ranks as the numbers of factor
columns of non-negligible energy. Using variational inference in the proposed
probabilistic model results in an iterative algorithm that comprises
closed-form updates. Its Bayesian nature completely avoids the ubiquitous in
regularization-based methods task of hyper-parameter tuning. Simulation results
with synthetic data are reported, which demonstrate the effectiveness of the
proposed scheme in terms of both rank estimation and model fitting.
| stat |
A stochastic alternating minimizing method for sparse phase retrieval | Sparse phase retrieval plays an important role in many fields of applied
science and thus attracts lots of attention. In this paper, we propose a
\underline{sto}chastic alte\underline{r}nating \underline{m}inimizing method
for \underline{sp}arse ph\underline{a}se \underline{r}etrieval
(\textit{StormSpar}) algorithm which {emprically} is able to recover
$n$-dimensional $s$-sparse signals from only $O(s\,\mathrm{log}\, n)$ number of
measurements without a desired initial value required by many existing methods.
In \textit{StormSpar}, the hard-thresholding pursuit (HTP) algorithm is
employed to solve the sparse constraint least square sub-problems. The main
competitive feature of \textit{StormSpar} is that it converges globally
requiring optimal order of number of samples with random initialization.
Extensive numerical experiments are given to validate the proposed algorithm.
| stat |
Certifiably Optimal Sparse Inverse Covariance Estimation | We consider the maximum likelihood estimation of sparse inverse covariance
matrices. We demonstrate that current heuristic approaches primarily encourage
robustness, instead of the desired sparsity. We give a novel approach that
solves the cardinality constrained likelihood problem to certifiable
optimality. The approach uses techniques from mixed-integer optimization and
convex optimization, and provides a high-quality solution with a guarantee on
its suboptimality, even if the algorithm is terminated early. Using a variety
of synthetic and real datasets, we demonstrate that our approach can solve
problems where the dimension of the inverse covariance matrix is up to 1,000s.
We also demonstrate that our approach produces significantly sparser solutions
than Glasso and other popular learning procedures, makes less false
discoveries, while still maintaining state-of-the-art accuracy.
| stat |
Kernel regression in high dimensions: Refined analysis beyond double
descent | In this paper, we provide a precise characterization of generalization
properties of high dimensional kernel ridge regression across the under- and
over-parameterized regimes, depending on whether the number of training data n
exceeds the feature dimension d. By establishing a bias-variance decomposition
of the expected excess risk, we show that, while the bias is (almost)
independent of d and monotonically decreases with n, the variance depends on n,
d and can be unimodal or monotonically decreasing under different
regularization schemes. Our refined analysis goes beyond the double descent
theory by showing that, depending on the data eigen-profile and the level of
regularization, the kernel regression risk curve can be a double-descent-like,
bell-shaped, or monotonic function of n. Experiments on synthetic and real data
are conducted to support our theoretical findings.
| stat |
Amortized variance reduction for doubly stochastic objectives | Approximate inference in complex probabilistic models such as deep Gaussian
processes requires the optimisation of doubly stochastic objective functions.
These objectives incorporate randomness both from mini-batch subsampling of the
data and from Monte Carlo estimation of expectations. If the gradient variance
is high, the stochastic optimisation problem becomes difficult with a slow rate
of convergence. Control variates can be used to reduce the variance, but past
approaches do not take into account how mini-batch stochasticity affects
sampling stochasticity, resulting in sub-optimal variance reduction. We propose
a new approach in which we use a recognition network to cheaply approximate the
optimal control variate for each mini-batch, with no additional model gradient
computations. We illustrate the properties of this proposal and test its
performance on logistic regression and deep Gaussian processes.
| stat |
Multidimensional Adaptive Penalised Splines with Application to Neurons'
Activity Studies | P-spline models have achieved great popularity both in statistical and in
applied research. A possible drawback of P-spline is that they assume a smooth
transition of the covariate effect across its whole domain. In some practical
applications, however, it is desirable and needed to adapt smoothness locally
to the data, and adaptive P-splines have been suggested. Yet, the extra
flexibility afforded by adaptive P-spline models is obtained at the cost of a
high computational burden, especially in a multidimensional setting.
Furthermore, to the best of our knowledge, the literature lacks proposals for
adaptive P-splines in more than two dimensions. Motivated by the need for
analysing data derived from experiments conducted to study neurons' activity in
the visual cortex, this work presents a novel locally adaptive anisotropic
P-spline model in two (e.g., space) and three (space and time) dimensions.
Estimation is based on the recently proposed SOP (Separation of Overlapping
Precision matrices) method, which provides the speed we look for. The practical
performance of the proposal is evaluated through simulations, and comparisons
with alternative methods are reported. In addition to the spatio-temporal
analysis of the data that motivated this work, we also discuss an application
in two dimensions on the absenteeism of workers.
| stat |
No increase in COVID-19 mortality after the 2020 primary elections in
the USA | We examined the impact of voting on the spread of COVID-19 after the US
primary elections held from March 17 to July 11, 2020 (1574 counties across 34
states). We estimated the average effect of treatment on the treated (ATT)
using a non-parametric, generalized difference-in-difference estimator with a
matching procedure for panel data. Separately, we estimated the time-varying
reproduction number $R_t$ using a semi-mechanistic Bayesian hierarchical model
at the state level. We found no evidence of a spike in COVID-19 deaths in the
period immediately following the primaries. It is possible that elections can
be held safely, without necessarily contributing to spreading the epidemic.
Appropriate precautionary measures that enforce physical distancing and
mask-wearing are likely needed during general elections, with larger turnout or
colder weather.
| stat |
Zoetrope Genetic Programming for Regression | The Zoetrope Genetic Programming (ZGP) algorithm is based on an original
representation for mathematical expressions, targeting evolutionary symbolic
regression.The zoetropic representation uses repeated fusion operations between
partial expressions, starting from the terminal set. Repeated fusions within an
individual gradually generate more complex expressions, ending up in what can
be viewed as new features. These features are then linearly combined to best
fit the training data. ZGP individuals then undergo specific crossover and
mutation operators, and selection takes place between parents and offspring.
ZGP is validated using a large number of public domain regression datasets, and
compared to other symbolic regression algorithms, as well as to traditional
machine learning algorithms. ZGP reaches state-of-the-art performance with
respect to both types of algorithms, and demonstrates a low computational time
compared to other symbolic regression approaches.
| stat |
A Review of Stochastic Block Models and Extensions for Graph Clustering | There have been rapid developments in model-based clustering of graphs, also
known as block modelling, over the last ten years or so. We review different
approaches and extensions proposed for different aspects in this area, such as
the type of the graph, the clustering approach, the inference approach, and
whether the number of groups is selected or estimated. We also review models
that combine block modelling with topic modelling and/or longitudinal
modelling, regarding how these models deal with multiple types of data. How
different approaches cope with various issues will be summarised and compared,
to facilitate the demand of practitioners for a concise overview of the current
status of these areas of literature.
| stat |
Sharp convergence rates for Langevin dynamics in the nonconvex setting | We study the problem of sampling from a distribution $p^*(x) \propto
\exp\left(-U(x)\right)$, where the function $U$ is $L$-smooth everywhere and
$m$-strongly convex outside a ball of radius $R$, but potentially nonconvex
inside this ball. We study both overdamped and underdamped Langevin MCMC and
establish upper bounds on the number of steps required to obtain a sample from
a distribution that is within $\epsilon$ of $p^*$ in $1$-Wasserstein distance.
For the first-order method (overdamped Langevin MCMC), the iteration complexity
is $\tilde{\mathcal{O}}\left(e^{cLR^2}d/\epsilon^2\right)$, where $d$ is the
dimension of the underlying space. For the second-order method (underdamped
Langevin MCMC), the iteration complexity is
$\tilde{\mathcal{O}}\left(e^{cLR^2}\sqrt{d}/\epsilon\right)$ for an explicit
positive constant $c$. Surprisingly, the iteration complexity for both these
algorithms is only polynomial in the dimension $d$ and the target accuracy
$\epsilon$. It is exponential, however, in the problem parameter $LR^2$, which
is a measure of non-log-concavity of the target distribution.
| stat |
Incremental Principal Component Analysis Exact implementation and
continuity corrections | This paper describes some applications of an incremental implementation of
the principal component analysis (PCA). The algorithm updates the
transformation coefficients matrix on-line for each new sample, without the
need to keep all the samples in memory. The algorithm is formally equivalent to
the usual batch version, in the sense that given a sample set the
transformation coefficients at the end of the process are the same. The
implications of applying the PCA in real time are discussed with the help of
data analysis examples. In particular we focus on the problem of the continuity
of the PCs during an on-line analysis.
| stat |
Geometrically Enriched Latent Spaces | A common assumption in generative models is that the generator immerses the
latent space into a Euclidean ambient space. Instead, we consider the ambient
space to be a Riemannian manifold, which allows for encoding domain knowledge
through the associated Riemannian metric. Shortest paths can then be defined
accordingly in the latent space to both follow the learned manifold and respect
the ambient geometry. Through careful design of the ambient metric we can
ensure that shortest paths are well-behaved even for deterministic generators
that otherwise would exhibit a misleading bias. Experimentally we show that our
approach improves interpretability of learned representations both using
stochastic and deterministic generators.
| stat |
Causal models for dynamical systems | A probabilistic model describes a system in its observational state. In many
situations, however, we are interested in the system's response under
interventions. The class of structural causal models provides a language that
allows us to model the behaviour under interventions. It can been taken as a
starting point to answer a plethora of causal questions, including the
identification of causal effects or causal structure learning. In this chapter,
we provide a natural and straight-forward extension of this concept to
dynamical systems, focusing on continuous time models. In particular, we
introduce two types of causal kinetic models that differ in how the randomness
enters into the model: it may either be considered as observational noise or as
systematic driving noise. In both cases, we define interventions and therefore
provide a possible starting point for causal inference. In this sense, the book
chapter provides more questions than answers. The focus of the proposed causal
kinetic models lies on the dynamics themselves rather than corresponding
stationary distributions, for example. We believe that this is beneficial when
the aim is to model the full time evolution of the system and data are measured
at different time points. Under this focus, it is natural to consider
interventions in the differential equations themselves.
| stat |
Unbiased Markov chain Monte Carlo with couplings | Markov chain Monte Carlo (MCMC) methods provide consistent of integrals as
the number of iterations goes to infinity. MCMC estimators are generally biased
after any fixed number of iterations. We propose to remove this bias by using
couplings of Markov chains together with a telescopic sum argument of Glynn and
Rhee (2014). The resulting unbiased estimators can be computed independently in
parallel. We discuss practical couplings for popular MCMC algorithms. We
establish the theoretical validity of the proposed estimators and study their
efficiency relative to the underlying MCMC algorithms. Finally, we illustrate
the performance and limitations of the method on toy examples, on an Ising
model around its critical temperature, on a high-dimensional variable selection
problem, and on an approximation of the cut distribution arising in Bayesian
inference for models made of multiple modules.
| stat |
AutoEncoder for Interpolation | In physical science, sensor data are collected over time to produce
timeseries data. However, depending on the real-world condition and underlying
physics of the sensor, data might be noisy. Besides, the limitation of
sample-time on sensors may not allow collecting data over all the timepoints,
may require some form of interpolation. Interpolation may not be smooth enough,
fail to denoise data, and derivative operation on noisy sensor data may be poor
that do not reveal any high order dynamics. In this article, we propose to use
AutoEncoder to perform interpolation that also denoise data simultaneously. A
brief example using a real-world is also provided.
| stat |
Binary Classification as a Phase Separation Process | We propose a new binary classification model called Phase Separation Binary
Classifier (PSBC). It consists of a discretization of a nonlinear
reaction-diffusion equation coupled with an ODE, and is inspired by fluid
behavior, namely, on how binary fluids phase separate. Hence, parameters and
hyperparameters have physical meaning, whose effects are carefully studied in
several different scenarios. PSBC's coefficients are trainable weights, chosen
according to a minimization problem using Gradient Descent; optimization relies
on a classical Backpropagation with weight sharing. The model can be seen under
the framework of feedforward networks, and is endowed with a nonlinear
activation function that is linear in trainable weights but polynomial in other
variables, yielding a cost function that is also polynomial. In view of the
model's connection with ODEs and parabolic PDEs, forward propagation amounts to
an initial value problem. Thus, stability conditions are established using the
concept of Invariant regions. Interesting model compression properties are
thoroughly discussed. We illustrate the classifier's qualities by applying it
to the subset of numbers "0" and "1" of the classical MNIST database, where we
are able to discern individuals with more than 94\% accuracy, sometimes using
less only about 10\% of variables.
| stat |
Gaussian Process for Tomography | Tomographic reconstruction, despite its revolutionary impact on a wide range
of applications, suffers from its ill-posed nature in that there is no unique
solution because of limited and noisy measurements. Traditional
optimization-based reconstruction relies on regularization to address this
issue; however, it faces its own challenge because the type of regularizer and
choice of regularization parameter are a critical but difficult decision.
Moreover, traditional reconstruction yields point estimates for the
reconstruction with no further indication of the quality of the solution. In
this work we address these challenges by exploring Gaussian processes (GPs).
Our proposed GP approach yields not only the reconstructed object through the
posterior mean but also a quantification of the solution uncertainties through
the posterior covariance. Furthermore, we explore the flexibility of the GP
framework to provide a robust model of the information across various length
scales in the object, as well as the complex noise in the measurements. We
illustrate the proposed approach on both synthetic and real tomography images
and show its unique capability of uncertainty quantification in the presence of
various types of noise, as well as reconstruction comparison with existing
methods.
| stat |
Insights on Variance Estimation for Blocked and Matched Pairs Designs | Evaluating blocked randomized experiments from a potential outcomes
perspective has two primary branches of work. The first focuses on larger
blocks, with multiple treatment and control units in each block. The second
focuses on matched pairs, with a single treatment and control unit in each
block. These literatures not only provide different estimators for the standard
errors of the estimated average impact, but they are also built on different
sets of assumptions. Neither literature handles cases with blocks of varying
size that contain singleton treatment or control units, a case which can occur
in a variety of contexts, such as with different forms of matching or
post-stratification. In this paper, we reconcile the literatures by carefully
examining the performance of variance estimators under several different
frameworks. We then use these insights to derive novel variance estimators for
experiments containing blocks of different sizes.
| stat |
Scalable Algorithms for Large Competing Risks Data | This paper develops two orthogonal contributions to scalable sparse
regression for competing risks time-to-event data. First, we study and
accelerate the broken adaptive ridge method (BAR), an $\ell_0$-based
iteratively reweighted $\ell_2$-penalization algorithm that achieves sparsity
in its limit, in the context of the Fine-Gray (1999) proportional
subdistributional hazards (PSH) model. In particular, we derive a new algorithm
for BAR regression, named cycBAR, that performs cyclic update of each
coordinate using an explicit thresholding formula. The new cycBAR algorithm
effectively avoids fitting multiple reweighted $\ell_2$-penalizations and thus
yields impressive speedups over the original BAR algorithm. Second, we address
a pivotal computational issue related to fitting the PSH model. Specifically,
the computation costs of the log-pseudo likelihood and its derivatives for PSH
model grow at the rate of $O(n^2)$ with the sample size $n$ in current
implementations. We propose a novel forward-backward scan algorithm that
reduces the computation costs to $O(n)$. The proposed method applies to both
unpenalized and penalized estimation for the PSH model and has exhibited
drastic speedups over current implementations. Finally, combining the two
algorithms can yields $>1,000$ fold speedups over the original BAR algorithm.
Illustrations of the impressive scalability of our proposed algorithm for large
competing risks data are given using both simulations and a United States Renal
Data System data.
| stat |
Sequential Monte Carlo with transformations | This paper introduces methodology for performing Bayesian inference
sequentially on a sequence of posteriors on spaces of different dimensions. We
show how this may be achieved through the use of sequential Monte Carlo (SMC)
samplers (Del Moral et al., 2006, 2007), making use of the full flexibility of
this framework in order that the method is computationally efficient. In
particular, we introduce the innovation of using deterministic transformations
to move particles effectively between target distributions with different
dimensions. This approach, combined with adaptive methods, yields an extremely
flexible and general algorithm for Bayesian model comparison that is suitable
for use in applications where the acceptance rate in reversible jump Markov
chain Monte Carlo (RJMCMC) is low. We demonstrate this approach on the
well-studied problem of model comparison for mixture models, and for the novel
application of inferring coalescent trees sequentially, as data arrives.
| stat |
Bayesian item response models for citizen science ecological data | So-called 'citizen science' data elicited from crowds has become increasingly
popular in many fields including ecology. However, the quality of this
information is being frequently debated by many within the scientific
community. Therefore, modern citizen science implementations require measures
of the users' proficiency that account for the difficulty of the tasks. We
introduce a new methodological framework of item response and linear logistic
test models with application to citizen science data used in ecology research.
This approach accommodates spatial autocorrelation within the item difficulties
and produces relevant ecological measures of species and site-related
difficulties, discriminatory power and guessing behavior. These, along with
estimates of the subject abilities allow better management of these programs
and provide deeper insights. This paper also highlights the fit of item
response models to big data via divide-and-conquer. We found that the suggested
methods outperform the traditional item response models in terms of RMSE,
accuracy, and WAIC based on leave-one-out cross-validation on simulated and
empirical data. We present a comprehensive implementation using a case study of
species identification in the Serengeti, Tanzania. The R and Stan codes are
provided for full reproducibility. Multiple statistical illustrations and
visualizations are given which allow practitioners the extrapolation to a wide
range of citizen science ecological problems.
| stat |
Template Matching Route Classification | This paper details a route classification method for American football using
a template matching scheme that is quick and does not require manual labeling.
Pre-defined routes from a standard receiver route tree are aligned closely with
game routes in order to determine the closest match. Based on a test game with
manually labeled routes, the method achieves moderate success with an overall
accuracy of 72\% of the 232 routes labeled correctly.
| stat |
The Kernel Spatial Scan Statistic | Kulldorff's (1997) seminal paper on spatial scan statistics (SSS) has led to
many methods considering different regions of interest, different statistical
models, and different approximations while also having numerous applications in
epidemiology, environmental monitoring, and homeland security. SSS provides a
way to rigorously test for the existence of an anomaly and provide statistical
guarantees as to how "anomalous" that anomaly is. However, these methods rely
on defining specific regions where the spatial information a point contributes
is limited to binary 0 or 1, of either inside or outside the region, while in
reality anomalies will tend to follow smooth distributions with decaying
density further from an epicenter. In this work, we propose a method that
addresses this shortcoming through a continuous scan statistic that generalizes
SSS by allowing the point contribution to be defined by a kernel. We provide
extensive experimental and theoretical results that shows our methods can be
computed efficiently while providing high statistical power for detecting
anomalous regions.
| stat |
The Significance Filter, the Winner's Curse and the Need to Shrink | The "significance filter" refers to focusing exclusively on statistically
significant results. Since frequentist properties such as unbiasedness and
coverage are valid only before the data have been observed, there are no
guarantees if we condition on significance. In fact, the significance filter
leads to overestimation of the magnitude of the parameter, which has been
called the "winner's curse". It can also lead to undercoverage of the
confidence interval. Moreover, these problems become more severe if the power
is low. While these issues clearly deserve our attention, they have been
studied only informally and mathematical results are lacking. Here we study
them from the frequentist and the Bayesian perspective. We prove that the
relative bias of the magnitude is a decreasing function of the power and that
the usual confidence interval undercovers when the power is less than 50%. We
conclude that failure to apply the appropriate amount of shrinkage can lead to
misleading inferences.
| stat |
Gradient Descent Finds Global Minima for Generalizable Deep Neural
Networks of Practical Sizes | In this paper, we theoretically prove that gradient descent can find a global
minimum of non-convex optimization of all layers for nonlinear deep neural
networks of sizes commonly encountered in practice. The theory developed in
this paper only requires the practical degrees of over-parameterization unlike
previous theories. Our theory only requires the number of trainable parameters
to increase linearly as the number of training samples increases. This allows
the size of the deep neural networks to be consistent with practice and to be
several orders of magnitude smaller than that required by the previous
theories. Moreover, we prove that the linear increase of the size of the
network is the optimal rate and that it cannot be improved, except by a
logarithmic factor. Furthermore, deep neural networks with the trainability
guarantee are shown to generalize well to unseen test samples with a natural
dataset but not a random dataset.
| stat |
Bayesian Functional Registration of fMRI Data | Functional magnetic resonance imaging (fMRI) has provided invaluable insight
into our understanding of human behavior. However, large inter-individual
differences in both brain anatomy and functional localization after anatomical
alignment remain a major limitation in conducting group analyses and performing
population-level inference. This paper addresses this problem by developing and
validating a new computational technique for reducing misalignment across
individuals in functional brain systems by spatially transforming each
subject's functional data to a common reference map. Our proposed Bayesian
functional registration approach allows us to assess differences in brain
function across subjects and individual differences in activation topology. It
combines intensity-based and feature-based information into an integrated
framework and allows inference to be performed on the transformation via the
posterior samples. We evaluate the method in a simulation study and apply it to
data from a study of thermal pain. We find that the proposed approach provides
increased sensitivity for group-level inference.
| stat |
Spying on the prior of the number of data clusters and the partition
distribution in Bayesian cluster analysis | Mixture models represent the key modelling approach for Bayesian cluster
analysis. Different likelihood and prior specifications are required to capture
the prototypical shape of the clusters. In addition, the mixture modelling
approaches also crucially differ in the specification of the prior on the
number of components and the prior on the component weight distribution. We
investigate how these specifications impact on the implicitly induced prior on
the number of 'filled' components, i.e., data clusters, and the prior on the
partitions. We derive computationally feasible calculations to obtain these
implicit priors for reasonable data analysis settings and make a reference
implementation available in the R package 'fipp'. In many applications the
implicit priors are of more practical relevance than the explicit priors
imposed and thus suitable prior specifications depend on the implicit priors
induced. We highlight the insights which may be gained from inspecting these
implicit priors by analysing them for three different modelling approaches
previously proposed for Bayesian cluster analysis. These modelling approaches
consist of the Dirichlet process mixture and the static and dynamic mixture of
finite mixtures model. The default priors suggested in the literature for these
modelling approaches are used and the induced priors compared. Based on the
implicit priors, we discuss the suitability of these modelling approaches and
prior specifications when aiming at sparse cluster solutions and flexibility in
the prior on the partitions.
| stat |
Algorithmic Bias in Recidivism Prediction: A Causal Perspective | ProPublica's analysis of recidivism predictions produced by Correctional
Offender Management Profiling for Alternative Sanctions (COMPAS) software tool
for the task, has shown that the predictions were racially biased against
African American defendants. We analyze the COMPAS data using a causal
reformulation of the underlying algorithmic fairness problem. Specifically, we
assess whether COMPAS exhibits racial bias against African American defendants
using FACT, a recently introduced causality grounded measure of algorithmic
fairness. We use the Neyman-Rubin potential outcomes framework for causal
inference from observational data to estimate FACT from COMPAS data. Our
analysis offers strong evidence that COMPAS exhibits racial bias against
African American defendants. We further show that the FACT estimates from
COMPAS data are robust in the presence of unmeasured confounding.
| stat |
Re-Examining Linear Embeddings for High-Dimensional Bayesian
Optimization | Bayesian optimization (BO) is a popular approach to optimize
expensive-to-evaluate black-box functions. A significant challenge in BO is to
scale to high-dimensional parameter spaces while retaining sample efficiency. A
solution considered in existing literature is to embed the high-dimensional
space in a lower-dimensional manifold, often via a random linear embedding. In
this paper, we identify several crucial issues and misconceptions about the use
of linear embeddings for BO. We study the properties of linear embeddings from
the literature and show that some of the design choices in current approaches
adversely impact their performance. We show empirically that properly
addressing these issues significantly improves the efficacy of linear
embeddings for BO on a range of problems, including learning a gait policy for
robot locomotion.
| stat |
Estimating Concurrent Climate Extremes: A Conditional Approach | Simultaneous concurrence of extreme values across multiple climate variables
can result in large societal and environmental impacts. Therefore, there is
growing interest in understanding these concurrent extremes. In many
applications, not only the frequency but also the magnitude of concurrent
extremes are of interest. One way to approach this problem is to study the
distribution of one climate variable given that another is extreme. In this
work we develop a statistical framework for estimating bivariate concurrent
extremes via a conditional approach, where univariate extreme value modeling is
combined with dependence modeling of the conditional tail distribution using
techniques from quantile regression and extreme value analysis to quantify
concurrent extremes. We focus on the distribution of daily wind speed
conditioned on daily precipitation taking its seasonal maximum. The Canadian
Regional Climate Model large ensemble is used to assess the performance of the
proposed framework both via a simulation study with specified dependence
structure and via an analysis of the climate model-simulated dependence
structure.
| stat |
Multilevel Gibbs Sampling for Bayesian Regression | Bayesian regression remains a simple but effective tool based on Bayesian
inference techniques. For large-scale applications, with complicated posterior
distributions, Markov Chain Monte Carlo methods are applied. To improve the
well-known computational burden of Markov Chain Monte Carlo approach for
Bayesian regression, we developed a multilevel Gibbs sampler for Bayesian
regression of linear mixed models. The level hierarchy of data matrices is
created by clustering the features and/or samples of data matrices.
Additionally, the use of correlated samples is investigated for variance
reduction to improve the convergence of the Markov Chain. Testing on a diverse
set of data sets, speed-up is achieved for almost all of them without
significant loss in predictive performance.
| stat |
Using Weighted P-Values in Fisher's Method | Fisher's method prescribes a way to combine p-values from multiple
experiments into a single p-value. However, the original method can only
determine a combined p-value analytically if all constituent p-values are
weighted equally. Here we present, with proof, a method to combine p-values
with arbitrary weights.
| stat |
Autoregressive Energy Machines | Neural density estimators are flexible families of parametric models which
have seen widespread use in unsupervised machine learning in recent years.
Maximum-likelihood training typically dictates that these models be constrained
to specify an explicit density. However, this limitation can be overcome by
instead using a neural network to specify an energy function, or unnormalized
density, which can subsequently be normalized to obtain a valid distribution.
The challenge with this approach lies in accurately estimating the normalizing
constant of the high-dimensional energy function. We propose the Autoregressive
Energy Machine, an energy-based model which simultaneously learns an
unnormalized density and computes an importance-sampling estimate of the
normalizing constant for each conditional in an autoregressive decomposition.
The Autoregressive Energy Machine achieves state-of-the-art performance on a
suite of density-estimation tasks.
| stat |
Reconciling modern machine learning practice and the bias-variance
trade-off | Breakthroughs in machine learning are rapidly changing science and society,
yet our fundamental understanding of this technology has lagged far behind.
Indeed, one of the central tenets of the field, the bias-variance trade-off,
appears to be at odds with the observed behavior of methods used in the modern
machine learning practice. The bias-variance trade-off implies that a model
should balance under-fitting and over-fitting: rich enough to express
underlying structure in data, simple enough to avoid fitting spurious patterns.
However, in the modern practice, very rich models such as neural networks are
trained to exactly fit (i.e., interpolate) the data. Classically, such models
would be considered over-fit, and yet they often obtain high accuracy on test
data. This apparent contradiction has raised questions about the mathematical
foundations of machine learning and their relevance to practitioners.
In this paper, we reconcile the classical understanding and the modern
practice within a unified performance curve. This "double descent" curve
subsumes the textbook U-shaped bias-variance trade-off curve by showing how
increasing model capacity beyond the point of interpolation results in improved
performance. We provide evidence for the existence and ubiquity of double
descent for a wide spectrum of models and datasets, and we posit a mechanism
for its emergence. This connection between the performance and the structure of
machine learning models delineates the limits of classical analyses, and has
implications for both the theory and practice of machine learning.
| stat |
Observing water level extremes in the Mekong River Basin: The benefit of
long-repeat orbit missions in a multi-mission satellite altimetry approach | Single-mission altimetric water level observations of rivers are spatially
and temporally limited, and thus they are often unable to quantify the full
extent of extreme flood events. Moreover, only missions with a short-repeat
orbit, such as Envisat, Jason-2, or SARAL, could provide meaningful time series
of water level variations directly. However, long or non-repeat orbit missions
such as CryoSat-2 have a very dense spatial resolution under the trade-off of a
repeat time insufficient for time series extraction. Combining data from
multiple altimeter missions into a multi-mission product allows for increasing
the spatial and temporal resolution of the data. In this study, we combined
water level data from CryoSat-2 with various observations from other altimeter
missions in the Mekong River Basin between 2008 and 2016 into one multi-mission
water level time series using the approach of universal kriging. In contrast to
former multi-mission altimetry methods, this approach allows for the
incorporation of CryoSat-2 data as well as data from other long or non-repeat
orbit missions, such as Envisat-EM or SARAL-DP. Additionally, for the first
time, data from tributaries are incorporated. The multi-mission time series
including CryoSat-2 data adequately reflects the general inter-annual flood
behaviour and the extreme floodings in 2008 and 2011. It performs better than
single-mission time series or multi-mission time series based only on
short-repeat orbit data. The Probability of Detection of the floodings with the
multi-mission altimetry was around 80\% while Envisat and Jason-2
single-mission altimetry could only detect around 40% of the floodings
correctly. However, small flash floods still remain undetectable.
| stat |
Learning Latent Representations of Bank Customers With The Variational
Autoencoder | Learning data representations that reflect the customers' creditworthiness
can improve marketing campaigns, customer relationship management, data and
process management or the credit risk assessment in retail banks. In this
research, we adopt the Variational Autoencoder (VAE), which has the ability to
learn latent representations that contain useful information. We show that it
is possible to steer the latent representations in the latent space of the VAE
using the Weight of Evidence and forming a specific grouping of the data that
reflects the customers' creditworthiness. Our proposed method learns a latent
representation of the data, which shows a well-defied clustering structure
capturing the customers' creditworthiness. These clusters are well suited for
the aforementioned banks' activities. Further, our methodology generalizes to
new customers, captures high-dimensional and complex financial data, and scales
to large data sets.
| stat |
Consistency of semi-supervised learning algorithms on graphs: Probit and
one-hot methods | Graph-based semi-supervised learning is the problem of propagating labels
from a small number of labelled data points to a larger set of unlabelled data.
This paper is concerned with the consistency of optimization-based techniques
for such problems, in the limit where the labels have small noise and the
underlying unlabelled data is well clustered. We study graph-based probit for
binary classification, and a natural generalization of this method to
multi-class classification using one-hot encoding. The resulting objective
function to be optimized comprises the sum of a quadratic form defined through
a rational function of the graph Laplacian, involving only the unlabelled data,
and a fidelity term involving only the labelled data. The consistency analysis
sheds light on the choice of the rational function defining the optimization.
| stat |
Penalized robust estimators in logistic regression with applications to
sparse models | Sparse covariates are frequent in classification and regression problems and
in these settings the task of variable selection is usually of interest. As it
is well known, sparse statistical models correspond to situations where there
are only a small number of non--zero parameters and for that reason, they are
much easier to interpret than dense ones. In this paper, we focus on the
logistic regression model and our aim is to address robust and penalized
estimation for the regression parameter. We introduce a family of penalized
weighted $M-$type estimators for the logistic regression parameter that are
stable against atypical data. We explore different penalizations functions and
we introduce the so--called Sign penalization. This new penalty has the
advantage that it depends only on one penalty parameter, avoiding arbitrary
tuning constants. We discuss the variable selection capability of the given
proposals as well as their asymptotic behaviour. Through a numerical study, we
compare the finite sample performance of the proposal corresponding to
different penalized estimators either robust or classical, under different
scenarios. A robust cross--validation criterion is also presented. The analysis
of two real data sets enables to investigate the stability of the penalized
estimators to the presence of outliers.
| stat |
Cross-validation Confidence Intervals for Test Error | This work develops central limit theorems for cross-validation and consistent
estimators of its asymptotic variance under weak stability conditions on the
learning algorithm. Together, these results provide practical,
asymptotically-exact confidence intervals for $k$-fold test error and valid,
powerful hypothesis tests of whether one learning algorithm has smaller
$k$-fold test error than another. These results are also the first of their
kind for the popular choice of leave-one-out cross-validation. In our real-data
experiments with diverse learning algorithms, the resulting intervals and tests
outperform the most popular alternative methods from the literature.
| stat |
A novel discrete grey seasonal model and its applications | In order to accurately describe real systems with seasonal disturbances,
which normally appear monthly or quarterly cycles, a novel discrete grey
seasonal model, abbreviated as , is put forward by incorporating the seasonal
dummy variables into the conventional model. Moreover, the mechanism and
properties of this proposed model are discussed in depth, revealing the
inherent differences from the existing seasonal grey models. For validation and
explanation purposes, the proposed model is implemented to describe three
actual cases with monthly and quarterly seasonal fluctuations (quarterly wind
power production, quarterly PM10, and monthly natural gas consumption), in
comparison with five competing models involving grey prediction models ,
conventional econometric technology , and artificial intelligences .
Experimental results from the cases consistently demonstrated that the proposed
model significantly outperforms the other benchmark models in terms of several
error criteria. Moreover, further discussions about the influences of different
sequence lengths on the forecasting performance reveal that the proposed model
still performs the best with strong robustness and high reliability in
addressing seasonal sequences. In general, the new model is validated to be a
powerful and promising methodology for handling sequences with seasonal
fluctuations.
| stat |
Shift identification in time varying regression quantiles | This article investigates whether time-varying quantile regression curves are
the same up to the horizontal shift or not. The errors and the covariates
involved in the regression model are allowed to be locally stationary. We
formalise this issue in a corresponding non-parametric hypothesis testing
problem, and develop a integrated-squared-norm based test (SIT) as well as a
simultaneous confidence band (SCB) approach. The asymptotic properties of SIT
and SCB under null and local alternatives are derived. We then propose valid
wild bootstrap algorithms to implement SIT and SCB. Furthermore, the usefulness
of the proposed methodology is illustrated via analysing simulated and real
data related to Covid-19 outbreak and climate science.
| stat |
Identification, Doubly Robust Estimation, and Semiparametric Efficiency
Theory of Nonignorable Missing Data With a Shadow Variable | We consider identification and estimation with an outcome missing not at
random (MNAR). We study an identification strategy based on a so-called shadow
variable. A shadow variable is assumed to be correlated with the outcome, but
independent of the missingness process conditional on the outcome and fully
observed covariates. We describe a general condition for nonparametric
identification of the full data law under MNAR using a valid shadow variable.
Our condition is satisfied by many commonly-used models; moreover, it is
imposed on the complete cases, and therefore has testable implications with
observed data only. We describe semiparametric estimation methods and evaluate
their performance on both simulation data and a real data example. We
characterize the semiparametric efficiency bound for the class of regular and
asymptotically linear estimators, and derive a closed form for the efficient
influence function.
| stat |
Sampling designs for epidemic prevalence estimation | Intuitively, sampling is likely to be more efficient for prevalence
estimation, if the cases (or positives) have a relatively higher representation
in the sample than in the population. In case the virus is transmitted via
personal contacts, contact tracing of the observed cases (but not noncases), to
be referred to as \emph{adaptive network tracing}, can generate a higher yield
of cases than random sampling from the population. The efficacy of relevant
designs for cross-sectional and change estimation is investigated. The
availability of these designs allows one unite tracing for combating the
epidemic and sampling for estimating the prevalence in a single endeavour.
| stat |
Weighted Ensemble of Statistical Models | We present a detailed description of our submission for the M4 forecasting
competition, in which it ranked 3rd overall. Our solution utilizes several
commonly used statistical models, which are weighted according to their
performance on historical data. We cluster series within each type of frequency
with respect to the existence of trend and seasonality. Every class of series
is assigned a different set of models to combine. Combination weights are
chosen separately for each series. We conduct experiments with a holdout set to
manually pick pools of models that perform best for a given series type, as
well as to choose the combination approaches.
| stat |
How Close and How Much? Linking Health Outcomes to Built Environment
Spatial Distributions | Built environment features (BEFs) refer to aspects of the human constructed
environment, which may in turn support or restrict health related behaviors and
thus impact health. In this paper we are interested in understanding whether
the spatial distribution and quantity of fast food restaurants (FFRs) influence
the risk of obesity in schoolchildren. To achieve this goal, we propose a
two-stage Bayesian hierarchical modeling framework. In the first stage,
examining the position of FFRs relative to that of some reference locations -
in our case, schools - we model the distances of FFRs from these reference
locations as realizations of Inhomogenous Poisson processes (IPP). With the
goal of identifying representative spatial patterns of exposure to FFRs, we
model the intensity functions of the IPPs using a Bayesian non-parametric
viewpoint and specifying a Nested Dirichlet Process prior. The second stage
model relates exposure patterns to obesity, offering two different approaches
to accommodate uncertainty in the exposure patterns estimated in the first
stage: in the first approach the odds of obesity at the school level is
regressed on cluster indicators, each representing a major pattern of exposure
to FFRs. In the second, we employ Bayesian Kernel Machine regression to relate
the odds of obesity to the multivariate vector reporting the degree of
similarity of a given school to all other schools. Our analysis on the
influence of patterns of FFR occurrence on obesity among Californian
schoolchildren has indicated that, in 2010, among schools that are consistently
assigned to a cluster, there is a lower odds of obesity amongst 9th graders who
attend schools with most distant FFR occurrences in a 1-mile radius as compared
to others.
| stat |
Balancing covariates in randomized experiments with the Gram-Schmidt
Walk design | The design of experiments involves a compromise between covariate balance and
robustness. This paper introduces an experimental design that admits precise
control over this trade-off. The design is specified by a parameter that bounds
the worst-case mean square error of an estimator of the average treatment
effect. Subject to the experimenter's desired level of robustness, the design
aims to simultaneously balance all linear functions of the covariates. The
achieved level of balance is considerably better than what a fully random
assignment would produce, and it is close to optimal given the desired level of
robustness. We show that the mean square error of the estimator is bounded by
the minimum of the loss function of a ridge regression of the potential
outcomes on the covariates. One may thus interpret the approach as regression
adjustment by design. Finally, we provide non-asymptotic tail bounds for the
estimator, which facilitate the construction of conservative confidence
intervals.
| stat |
Tools for analyzing R code the tidy way | With the current emphasis on reproducibility and replicability, there is an
increasing need to examine how data analyses are conducted. In order to analyze
the between researcher variability in data analysis choices as well as the
aspects within the data analysis pipeline that contribute to the variability in
results, we have created two R packages: matahari and tidycode. These packages
build on methods created for natural language processing; rather than allowing
for the processing of natural language, we focus on R code as the substrate of
interest. The matahari package facilitates the logging of everything that is
typed in the R console or in an R script in a tidy data frame. The tidycode
package contains tools to allow for analyzing R calls in a tidy manner. We
demonstrate the utility of these packages as well as walk through two examples.
| stat |
Nonparametric C- and D-vine based quantile regression | Quantile regression is a field with steadily growing importance in
statistical modeling. It is a complementary method to linear regression, since
computing a range of conditional quantile functions provides a more accurate
modelling of the stochastic relationship among variables, especially in the
tails. We introduce a novel non-restrictive and highly flexible nonparametric
quantile regression approach based on C- and D-vine copulas. Vine copulas allow
for separate modeling of marginal distributions and the dependence structure in
the data, and can be expressed through a graph theoretical model given by a
sequence of trees. This way we obtain a quantile regression model, that
overcomes typical issues of quantile regression such as quantile crossings or
collinearity, the need for transformations and interactions of variables. Our
approach incorporates a two-step ahead ordering of variables, by maximizing the
conditional log-likelihood of the tree sequence, while taking into account the
next two tree levels. Further, we show that the nonparametric conditional
quantile estimator is consistent. The performance of the proposed methods is
evaluated in both low- and high-dimensional settings using simulated and real
world data. The results support the superior prediction ability of the proposed
models.
| stat |
Data-Driven Factor Graphs for Deep Symbol Detection | Many important schemes in signal processing and communications, ranging from
the BCJR algorithm to the Kalman filter, are instances of factor graph methods.
This family of algorithms is based on recursive message passing-based
computations carried out over graphical models, representing a factorization of
the underlying statistics. Consequently, in order to implement these
algorithms, one must have accurate knowledge of the statistical model of the
considered signals. In this work we propose to implement factor graph methods
in a data-driven manner. In particular, we propose to use machine learning (ML)
tools to learn the factor graph, instead of the overall system task, which in
turn is used for inference by message passing over the learned graph. We apply
the proposed approach to learn the factor graph representing a finite-memory
channel, demonstrating the resulting ability to implement BCJR detection in a
data-driven fashion. We demonstrate that the proposed system, referred to as
BCJRNet, learns to implement the BCJR algorithm from a small training set, and
that the resulting receiver exhibits improved robustness to inaccurate training
compared to the conventional channel-model-based receiver operating under the
same level of uncertainty. Our results indicate that by utilizing ML tools to
learn factor graphs from labeled data, one can implement a broad range of
model-based algorithms, which traditionally require full knowledge of the
underlying statistics, in a data-driven fashion.
| stat |
How much is optimal reinsurance degraded by error? | The literature on optimal reinsurance does not deal with how much the
effectiveness of such solutions is degraded by errors in parameters and models.
The issue is investigated through both asymptotics and numerical studies. It is
shown that the rate of degradation is often $O(1/n)$ as the sample size $n$ of
historical observations becomes infinite. Criteria based on Value at Risk are
exceptions that may achieve only $O(1/\sqrt{n})$. These theoretical results are
supported by numerical studies. A Bayesian perspective on how to integrate risk
caused by parameter error is offered as well.
| stat |
A multinomial Asymptotic Representation of Zenga's Discrete Index, its
Influence Function and Data-driven Applications | In this paper, we consider the Zenga index, one of the most recent inequality
index. We keep the finite-valued original form and address the asymptotic
theory. The asymptotic normality is established through a multinomial
representation. The Influence function is also given. Th results are simulated
and applied to Senegalese data.
| stat |
Relative Feature Importance | Interpretable Machine Learning (IML) methods are used to gain insight into
the relevance of a feature of interest for the performance of a model. Commonly
used IML methods differ in whether they consider features of interest in
isolation, e.g., Permutation Feature Importance (PFI), or in relation to all
remaining feature variables, e.g., Conditional Feature Importance (CFI). As
such, the perturbation mechanisms inherent to PFI and CFI represent extreme
reference points. We introduce Relative Feature Importance (RFI), a
generalization of PFI and CFI that allows for a more nuanced feature importance
computation beyond the PFI versus CFI dichotomy. With RFI, the importance of a
feature relative to any other subset of features can be assessed, including
variables that were not available at training time. We derive general
interpretation rules for RFI based on a detailed theoretical analysis of the
implications of relative feature relevance, and demonstrate the method's
usefulness on simulated examples.
| stat |
Hierarchical correction of $p$-values via a tree running
Ornstein-Uhlenbeck process | Statistical testing is classically used as an exploratory tool to search for
association between a phenotype and many possible explanatory variables. This
approach often leads to multiple dependence testing under dependence. We assume
a hierarchical structure between tests via an Ornstein-Uhlenbeckprocess on a
tree. The process correlation structure is used for smoothing the p-values. We
design a penalized estimation of the mean of the OU process for p-value
computation. The performances of the algorithm are assessed via simulations.
Its ability to discover new associations is demonstrated on a metagenomic
dataset. The corresponding R package is available from
https://github.com/abichat/zazou.
| stat |
Power Comparisons in 2x2 Contingency Tables: Odds Ratio versus Pearson
Correlation versus Canonical Correlation | It is an important inferential problem to test no association between two
binary variables based on data. Tests based on the sample odds ratio are
commonly used. We bring in a competing test based on the Pearson correlation
coefficient. In particular, the Odds ratio does not extend to higher order
contingency tables, whereas Pearson correlation does. It is important to
understand how Pearson correlation stacks against the odds ratio in 2x2 tables.
Another measure of association is the canonical correlation. In this paper, we
examine how competitive Pearson correlation is vis-\`a-vis odds ratio in terms
of power in the binary context, contrasting further with both the Wald Z and
Rao Score tests. We generated an extensive collection of joint distributions of
the binary variables and estimated the power of the tests under each joint
alternative distribution based on random samples. The consensus is that none of
the tests dominates the other.
| stat |
Context-dependent Ranking and Selection under a Bayesian Framework | We consider a context-dependent ranking and selection problem. The best
design is not universal but depends on the contexts. Under a Bayesian
framework, we develop a dynamic sampling scheme for context-dependent
optimization (DSCO) to efficiently learn and select the best designs in all
contexts. The proposed sampling scheme is proved to be consistent. Numerical
experiments show that the proposed sampling scheme significantly improves the
efficiency in context-dependent ranking and selection.
| stat |
Thompson Sampling via Local Uncertainty | Thompson sampling is an efficient algorithm for sequential decision making,
which exploits the posterior uncertainty to address the
exploration-exploitation dilemma. There has been significant recent interest in
integrating Bayesian neural networks into Thompson sampling. Most of these
methods rely on global variable uncertainty for exploration. In this paper, we
propose a new probabilistic modeling framework for Thompson sampling, where
local latent variable uncertainty is used to sample the mean reward.
Variational inference is used to approximate the posterior of the local
variable, and semi-implicit structure is further introduced to enhance its
expressiveness. Our experimental results on eight contextual bandit benchmark
datasets show that Thompson sampling guided by local uncertainty achieves
state-of-the-art performance while having low computational complexity.
| stat |
Copula Averaging for Tail Dependence in Insurance Claims Data | Analysing dependent risks is an important task for insurance companies. A
dependency is reflected in the fact that information about one random variable
provides information about the likely distribution of values of another random
variable. Insurance companies in particular must investigate such dependencies
between different lines of business and the effects that an extreme loss event,
such as an earthquake or hurricane, has across multiple lines of business
simultaneously. Copulas provide a popular model-based approach to analysing the
dependency between risks, and the coefficient of tail dependence is a measure
of dependence for extreme losses. Besides commonly used empirical estimators
for estimating the tail dependence coefficient, copula fitting can lead to
estimation of such coefficients directly or can verify their existence.
Generally, a range of copula models is available to fit a data set well,
leading to multiple different tail dependence results; a method based on
Bayesian model averaging is designed to obtain a unified estimate of tail
dependence. In this article, this model-based coefficient estimation method is
illustrated through a variety of copula fitting approaches and results are
presented for several simulated data sets and also a real general insurance
loss data set.
| stat |
The compound product distribution; a solution to the distributional
equation X=AX+1 | The solution of $ X=AX+1 $ is analyzed for a discrete variable $ A $ with $
\mathbb{P}\left[A=0\right]>0 $. Accordingly, a fast algorithm is presented to
calculate the obtained heavy tail density. To exemplify, the compound product
distribution is studied in detail for some particular families of
distributions.
| stat |
Reactive Sample Size for Heuristic Search in Simulation-based
Optimization | In simulation-based optimization, the optimal setting of the input parameters
of the objective function can be determined by heuristic optimization
techniques. However, when simulators model the stochasticity of real-world
problems, their output is a random variable and multiple evaluations of the
objective function are necessary to properly compare the expected performance
of different parameter settings. This paper presents a novel reactive sample
size algorithm based on parametric tests and indifference-zone selection, which
can be used for improving the efficiency and robustness of heuristic
optimization methods. The algorithm reactively decides, in an online manner,
the sample size to be used for each comparison during the optimization
according to observed statistical evidence. Tests employ benchmark functions
extended with artificial levels of noise and a simulation-based optimization
tool for hotel revenue management. Experimental results show that the reactive
method can improve the efficiency and robustness of simulation-based
optimization techniques.
| stat |
A unified framework for weighted parametric group sequential design
(WPGSD) | Group sequential design (GSD) is widely used in clinical trials in which
correlated tests of multiple hypotheses are used. Multiple primary objectives
resulting in tests with known correlations include evaluating 1) multiple
experimental treatment arms, 2) multiple populations, 3) the combination of
multiple arms and multiple populations, or 4) any asymptotically multivariate
normal tests. In this paper, we focus on the first 3 of these and extend the
framework of the weighted parametric multiple test procedure from fixed designs
with a single analysis per objective to a GSD setting where different
objectives may be assessed at the same or different times, each in a group
sequential fashion. Pragmatic methods for design and analysis of weighted
parametric group sequential design(WPGSD) under closed testing procedures are
proposed to maintain the strong control of familywise Type I error rate (FWER)
when correlations between tests are incorporated. This results in the ability
to relax testing bounds compared to designs not fully adjusting for known
correlations, increasing power or allowing decreased sample size. We illustrate
the proposed methods using clinical trial examples and conduct a simulation
study to evaluate the operating characteristics.
| stat |
A pseudo-likelihood approach for multivariate meta-analysis of test
accuracy studies with multiple thresholds | Multivariate meta-analysis of test accuracy studies when tests are evaluated
in terms of sensitivity and specificity at more than one threshold represents
an effective way to synthesize results by fully exploiting the data, if
compared to univariate meta-analyses performed at each threshold independently.
The approximation of logit transformations of sensitivities and specificities
at different thresholds through a normal multivariate random-effects model is a
recent proposal, that straightforwardly extends the bivariate models well
recommended for the one threshold case. However, drawbacks of the approach,
such as poor estimation of the within-study correlations between sensitivities
and between specificities and severe computational issues, can make it
unappealing. We propose an alternative method for inference on common
diagnostic measures using a pseudo-likelihood constructed under a working
independence assumption between sensitivities and between specificities at
different thresholds in the same study. The method does not require
within-study correlations, overcomes the convergence issues and can be
effortlessly implemented. Simulation studies highlight a satisfactory
performance of the method, remarkably improving the results from the
multivariate normal counterpart under different scenarios. The
pseudo-likelihood approach is illustrated in the evaluation of a test used for
diagnosis of pre-eclampsia as a cause of maternal and perinatal morbidity and
mortality.
| stat |
Gaussian Process Optimization with Adaptive Sketching: Scalable and No
Regret | Gaussian processes (GP) are a well studied Bayesian approach for the
optimization of black-box functions. Despite their effectiveness in simple
problems, GP-based algorithms hardly scale to high-dimensional functions, as
their per-iteration time and space cost is at least quadratic in the number of
dimensions $d$ and iterations $t$. Given a set of $A$ alternatives to choose
from, the overall runtime $O(t^3A)$ is prohibitive. In this paper we introduce
BKB (budgeted kernelized bandit), a new approximate GP algorithm for
optimization under bandit feedback that achieves near-optimal regret (and hence
near-optimal convergence rate) with near-constant per-iteration complexity and
remarkably no assumption on the input space or covariance of the GP.
We combine a kernelized linear bandit algorithm (GP-UCB) with randomized
matrix sketching based on leverage score sampling, and we prove that randomly
sampling inducing points based on their posterior variance gives an accurate
low-rank approximation of the GP, preserving variance estimates and confidence
intervals. As a consequence, BKB does not suffer from variance starvation, an
important problem faced by many previous sparse GP approximations. Moreover, we
show that our procedure selects at most $\tilde{O}(d_{eff})$ points, where
$d_{eff}$ is the effective dimension of the explored space, which is typically
much smaller than both $d$ and $t$. This greatly reduces the dimensionality of
the problem, thus leading to a $O(TAd_{eff}^2)$ runtime and $O(A d_{eff})$
space complexity.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.