Search

Distributional regression models that overcome the traditional focus on relating the conditional mean of the response to explanatory variables and instead target either the complete conditional response distribution or more general features thereof have seen increasing interest in the past decade. The current state of distributional regression will be discussed, with a particular focus on the four most prominent model classes: (i) generalized additive models for location, scale and shape, (ii) conditional transformation models and distribution regression, (iii) density regression, and (iv) quantile and expectile regression. Characteristics of the different distributional regression approaches will be provided to establish a structured overview on the similarities and differences with respect to the required assumptions on the conditional response distribution, theoretical properties, and the availability of software implementations. In addition, challenges arising in the interpretability of distributional regression models will be discussed and all four approaches will be illustrated with an application analyzing determinants of income distributions from the German Socio-Economic Panel (GSOEP).

The Vector AutoRegressive Moving Average (VARMA) model is fundamental to the theory of multivariate time series; however, identifiability issues have led practitioners to abandon it in favor of the simpler but more restrictive Vector AutoRegressive (VAR) model. We narrow this gap with a new optimization-based approach to VARMA identification built upon the principle of parsimony. Among all equivalent data-generating models, we use convex optimization to seek the parameterization that is "simplest" in a certain sense. A user-specified strongly convex penalty is used to measure model simplicity, and that same penalty is then used to define an estimator that can be efficiently computed. We establish consistency of our estimators in a double-asymptotic regime. Our non-asymptotic error bound analysis accommodates both model specification and parameter estimation steps, a feature that is crucial for studying large-scale VARMA algorithms. Our analysis also provides new results on penalized estimation of infinite-order VAR, and elastic net regression under a singular covariance structure of regressors, which may be of independent interest. We illustrate the advantage of our method over VAR alternatives on three real data examples.

The randomized treatment allocation process in a response adaptive clinical trial is formulated as a stochastic sequential decision problem and an algorithm is proposed to approximate the optimal value under the average reward criterion. When the information of previous treatment allocations and associated responses are summarized with sufficient statistics for unknown parameters, the decision process becomes a Markov process, on which a span-contractor operator is defined. It is proven that the average reward under the policy identified from the span-contractor operator converges almost surely to the optimal value. Numerical results reveal that the sequential procedure based on the controlled Markov process shows superior ethical advantage and at the same time produces good statistical power for large sample sizes such as 200 or larger.

Missing values in data impede the task of inference for population parameters of interest. Multiple Imputation (MI) is a popular method for handling missing data since it accounts for the uncertainty of missing values. Inference in MI involves combining point and variance estimates from each imputed dataset via Rubin’s rules. A sufficient condition for these rules is that the estimator is approximately (multivariate) normally distributed. However, these traditional combining rules get computationally cumbersome for multicomponent parameters of interest, and unreliable at high rates of missingness (due to an unstable variance matrix). New combining rules for univariate F- and Beta-statistics from multiply-imputed data are proposed for decisions about multicomponent parameters. The proposed combining rules have the advantage of being computationally convenient since they only involve univariate F- and Beta-statistics, while providing the same inferential reliability as the traditional multivariate combining rules. Simulation study is conducted to demonstrate that the proposed method has good statistical properties of maintaining low type I and type II error rates at relatively large proportions of missingness. The general applicability of the proposed method is demonstrated within a lead exposure study to assess the association between lead exposure and neurological motor function.

For each of (i) arbitrary stochastic reset, (ii) deterministic reset with arbitrary period, (iii) reset at arbitrary constant rate, and then in the sense of either (a) first-order stochastic dominance or (b) expectation (i.e. for each of the six possible combinations of the preceding), those laws of random times are precisely characterized that are rendered no bigger [rendered no smaller; left invariant] by all possible restart laws (within the classes (i), (ii), (iii), as the case may be). Partial results in the same vein for reset with branching are obtained. In particular it is found that deterministic and arbitrary stochastic restart lead to the same characterizations, but this equivalence fails to persist for exponential (constant-rate) reset.

An approach to modelling volatile financial return series using stationary d-vine copula processes combined with Lebesgue-measure-preserving transformations known as v-transforms is proposed. By developing a method of stochastically inverting v-transforms, models are constructed that can describe both stochastic volatility in the magnitude of price movements and serial correlation in their directions. In combination with parametric marginal distributions it is shown that these models can rival and sometimes outperform well-known models in the extended GARCH family. 1

Asymptotic properties of classical kernel estimators for the spectral density are studied in the locally stationary framework. In particular, it is shown that for a locally stationary process standard spectral density estimators consistently estimate the time-averaged spectral density. This result is complemented by some illustrative examples and applications including HAC-inference in the multiple linear regression model, a simple visual tool for the detection of unconditional heteroskedasticity and a test for covariance stationarity.

A stochastic calculus is given for processes described by stochastic integrals with respect to fractional Brownian motions and Rosenblatt processes somewhat analogous to the stochastic calculus for Ito processes. These processes for this stochastic calculus arise naturally from a stochastic chain rule for functionals of Rosenblatt processes; and some Ito-type expressions are given here. Furthermore, there is some analysis of these results for their applications to problems using Rosenblatt noise.

In this paper, we study the asymptotic behavior of the quadratic variation for the class of AR(1) processes driven by white noise in the second Wiener chaos. Using tools from the analysis on Wiener space, we give an upper bound for the total-variation speed of convergence to the normal law, which we apply to study the estimation of the model’s mean-reversion. Simulations are performed to illustrate the theoretical results.

We study the optimal sustainable harvesting of a population that lives in a random environment. The novelty of our setting is that we maximize the asymptotic harvesting yield, both in an expected value and almost sure sense, for a large class of harvesting strategies and unstructured population models. We prove under relatively weak assumptions that there exists a unique optimal harvesting strategy characterized by an optimal threshold below which the population is maintained at all times by utilizing a local time push-type policy. We also discuss, through Abelian limits, how our results are related to the optimal harvesting strategies when one maximizes the expected cumulative present value of the harvesting yield and establish a simple connection and ordering between the values and optimal boundaries. Finally, we explicitly characterize the optimal harvesting strategies in two different cases, one of which is the celebrated stochastic Verhulst Pearl logistic model of population growth.

We consider a Bessel process X of dimension δ ∈ ( 0 , 2 ) having 0 as a slowly reflecting (sticky) boundary point with a stickiness parameter 1 ∕ μ ∈ ( 0 , ∞ ) . We show that (i) the process X can be characterised through its square Y = X 2 as a unique weak solution to the SDE system d Y t = δ I ( Y t > 0 ) d t + 2 Y t d B t I ( Y t = 0 ) d t = 1 2 μ d l t 0 ( Y ) where B is a standard Brownian motion and l 0 ( Y ) is a diffusion local time process of Y at 0, and (ii) the transition density function of X can be expressed in the closed form as a convolution integral involving a Mittag-Leffler function and a modified Bessel function of the second kind. Appearance of the Mittag-Leffler function is novel in this context. We determine a (sticky) boundary condition at zero that characterises the transition density function of X as a unique solution to the Kolmogorov backward/forward equation of X . We also show that the convolution integral can be characterised as a unique solution to the generalised Abel equation of the second kind. Letting μ ↓ 0 (absorption) and μ ↑ ∞ (instantaneous reflection) the closed-form expression for the transition density function of X reduces to the ones found by Feller (1951) and Molchanov (1967) respectively.

High-dimensional linear regression has been intensively studied in the community of statistics in the last two decades. For the convenience of theoretical analyses, classical methods usually assume independent observations and sub-Gaussian-tailed errors. However, neither of them hold in many real high-dimensional time-series data. Recently [Sun, Zhou, Fan, 2019, J. Amer. Stat. Assoc., in press] proposed Adaptive Huber Regression (AHR) to address the issue of heavy-tailed errors. They discover that the robustification parameter of the Huber loss should adapt to the sample size, the dimensionality, and the moments of the heavy-tailed errors. We progress in a vertical direction and justify AHR on dependent observations. Specifically, we consider an important dependence structure -- Markov dependence. Our results show that the Markov dependence impacts on the adaption of the robustification parameter and the estimation of regression coefficients in the way that the sample size should be discounted by a factor depending on the spectral gap of the underlying Markov chain.

This paper answers a long-standing open question concerning the 1 ∕ e -strategy for the problem of best choice. N candidates for a job arrive at times independently uniformly distributed in [ 0 , 1 ] . The interviewer knows how each candidate ranks relative to all others seen so far, and must immediately appoint or reject each candidate as they arrive. The aim is to choose the best overall. The 1 ∕ e strategy is to follow the rule: ‘Do nothing until time 1 ∕ e , then appoint the first candidate thereafter who is best so far (if any).’ The question, first discussed with Larry Shepp in 1983, was to know whether the 1 ∕ e -strategy is optimal if one has ‘no information about the total number of options’. Quite what this might mean is open to various interpretations, but we shall take the proportional-increment process formulation of Bruss and Yor (2012). Such processes are shown to have a very rigid structure, being time-changed pure birth processes, and this allows some precise distributional calculations, from which we deduce that the 1 ∕ e -strategy is in fact not optimal.

On being told that a piece of work he thought was his discovery had duplicated an earlier mathematician’s work, Larry Shepp once replied “Yes, but when I discovered it, it stayed discovered.” In this spirit we give discussion and probabilistic proofs of two related known results (Moon 1963, Joe 1988) on random tournaments which seem surprisingly unknown to modern probabilists. In particular, our proof of Moon’s theorem on mean score sequences seems more constructive than previous proofs. This provides a comparatively concrete introduction to a longstanding mystery, the lack of a canonical construction for a joint distribution in the representation theorem for convex order.

Motivated in part by a problem in simulated tempering (a form of Markov chain Monte Carlo) we seek to minimise, in a suitable sense, the time it takes a (regular) diffusion with instantaneous reflection at 0 and 1 to travel from the origin to 1 and then return (the so-called commute time from 0 to 1). We consider the static and dynamic versions of this problem where the control mechanism is related to the diffusion’s drift via the corresponding scale function. In the static version the diffusion’s drift can be chosen at each point in [0,1], whereas in the dynamic version we are only able to choose the drift at each point at the time of first visiting that point. The dynamic version leads to a novel type of stochastic control problem.

Summary Clinical trials often aim to compare two groups of patients for efficacy and/or toxicity depending on covariates such as dose. Examples include the comparison of populations from different geographic regions or age classes or, alternatively, of different treatment groups. Similarity of these groups can be claimed if the difference in average outcome is below a certain margin over the entire covariate range. In this article, we consider the problem of testing for similarity in the case that efficacy and toxicity are measured as binary outcome variables. We develop a new test for the assessment of similarity of two groups for a single binary endpoint. Our approach is based on estimating the maximal deviation between the curves describing the responses of the two groups, followed by a parametric bootstrap test. Further, using a two-dimensional Gumbel-type model we develop methodology to establish similarity for (correlated) binary efficacy–toxicity outcomes. We investigate the operating characteristics of the proposed methodology by means of a simulation study and present a case study as an illustration.

A fixed-effects logit model that accounts for feedback effects of the dependent variable on the covariates is proposed. The model is formulated by including leads of the predetermined covariates among the regressors and it is proved to satisfy certain theoretical properties under some regularity conditions on the distribution of the covariates. Estimation is based on the Conditional Maximum Likelihood ( cml ) method for the static logit model and the Pseudo- cml ( pcml ) method for the corresponding dynamic formulation. Both methods have good finite-sample properties even when the required regularity conditions are not satisfied. An application is provided about female labor supply where we jointly account for the predetermined number of children and husbands’ income. Differently from previous studies, it emerges that female employment history does not affect future fertility choices and the husband’s earnings, as no evidence of feedback effects is found.

Karunanayaka and Tang (2017) introduced a class of compromise designs for estimating main effects under the baseline parameterization. Their approach is to add some runs to the basic one-factor-at-a-time design, and its implementation requires computer search except for the case of adding one run where a theoretical result is available. The reliance on computer search only allowed them to find the limited results from adding up to four runs to the basic one-factor-at-a-time design. In this paper, we provide a systematic construction of compromise designs, without computer search and without restriction on the run size and the number of factors. Closed-form expressions of the bias and efficiency criteria are obtained for the new designs. Both theoretical and empirical studies show that our designs outperform those of Karunanayaka and Tang (2017) saving small-sized problems.

Diurnal fluctuations in volatility are a well-documented stylized fact of intraday price data. This warrants an investigation how this intraday periodicity (IP) affects both finite sample as well as asymptotic properties of several popular realized estimators of daily integrated volatility which are based on functionals of a finite number of intraday returns. It turns out that most of the estimators considered in this study exhibit a finite-sample bias due to IP, which can however get negligible when the number of intraday returns diverges to infinity. The appropriate correction factors for this bias are derived based on estimates of the IP. The adequacy of the new corrections is evaluated by means of a Monte Carlo simulation study and an empirical example.

In this paper, we investigate temporal clusters of extremes defined as subsequent exceedances of high thresholds in a stationary time series. Two meaningful features of these clusters are the probability distribution of the cluster size and the ordinal patterns giving the relative positions of the data points within a cluster. Since these patterns take only the ordinal structure of consecutive data points into account, the method is robust under monotone transformations and measurement errors. We verify the existence of the corresponding limit distributions in the framework of regularly varying time series, develop non-parametric estimators and show their asymptotic normality under appropriate mixing conditions. The performance of the estimators is demonstrated in a simulated example and a real data application to discharge data of the river Rhine., Deutsche Forschungsgemeinschaft, Projekt DEAL

Predicting large fetuses at birth is of great interest to obstetricians. Using an NICHD Scandinavian Study that collected longitudinal ultrasound examination data during pregnancy, we estimate diagnostic accuracy parameters of estimated fetal weight (EFW) at various times during pregnancy in predicting large-for-gestational-age. We adopt a placement value based Bayesian regression model with random effects to estimate ROC curves. The use of placement values allows us to model covariate effects directly on the ROC curves and the adoption of a Bayesian approach accommodates the a priori constraint that an ROC curve of EFW near delivery should dominate another further away. The proposed methodology is shown to perform better than some alternative approaches in simulations and its application to the Scandinavian Study data suggests that diagnostic accuracy of EFW can improve about 65% from week 17 to 37 of gestation.

Ensuring aggregate food price stability requires a forward-looking assessment of the risk that unexpected deviations in individual food items’ inflation lead to large shocks in the aggregate food price inflation. To do so, we propose using a multivariate GARCH framework in combination with the Euler method to (1) estimate the conditional standard deviation and quantiles of the food price inflation shocks and (2) attribute the total risk to the underlying food items. For the FAO food price index, we find that even though meat inflation systematically has the highest weight in the aggregate index, cereal inflation is the main contributor to the total food price inflation risk over the period 1990–2018. The use of time series models and the Cornish-Fisher expansion make the risk characterization forward-looking and a potentially helpful tool for risk management.

The new model class of mixtures of generalised nonlinear models (GNMs) is introduced. The model is specified, identifiability issues discussed, the fitting in a maximum likelihood framework using the expectation-maximisation (EM) algorithm outlined and an appropriate computational implementation introduced. The new model class is applied to capture cross-country heterogeneity when considering the augmented Solow model including human capital accumulation as underlying model structure. The inherent heterogeneity is attributed to multiple regimes being present within the selected country data set. The results highlight that country-specific differences lead to distinct components. Countries belonging to the same component exhibit convergence to a homogeneous steady state. The components differ in the initial technological endowment and the contribution of the economic variables to economic growth.

Passes are undoubtedly the more frequent events in football and other team sports. Passing networks and their structural features can be useful to evaluate the style of play in terms of passing behavior, analyzing and quantifying interactions among players. The present paper aims to show how information retrieved from passing networks can have a relevant impact on predicting the match outcome. In particular, we focus on modeling both the scored goals by two competing teams and the goal difference between them. With this purpose, we fit these outcomes using Bayesian hierarchical models, including both in-match and network-based covariates to cover many aspects of the offensive actions on the pitch. Furthermore, we review and compare different approaches to include covariates in modeling football outcomes. The presented methodology is applied to a real dataset containing information on 125 matches of the 2016–2017 UEFA Champions League, involving 32 among the best European teams. From our results, shots on target, corners, and such passing network indicators are the main determinants of the considered football outcomes.

21 pages, 6 figures, 8 tables.-- This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License, Testing whether two spatial point processes have the same spatial distribution is an important task that can be addressed from different perspectives. A Kolmogorov-Smirnov test with asymptotic calibration and a Cramer von Mises type test with bootstrap calibration have recently been developed to compare the first-order intensity of two observed patterns. Motivated by common practice in epidemiological studies, we introduce a regression test based on the relative risk function with two alternative bootstrap calibrations. This paper compares the performance of these nonparametric tests through both an intensive simulation study, and the application to wildfire and crime data. The three tests provide good calibrations of the null hypothesis for simulated Poisson and non-Poisson spatial point processes, but the Cramer von Mises and regression tests outperform the cost-efficient Kolmogorov-Smirnov test in terms of power. In the real data analysis we have seen that the Kolmogorov-Smirnov test does not detect differences between spatial point patterns when dealing with sparse data. In view of these results, it would be preferable using the Cramer von Mises or regression tests despite their higher computational demand, This work has been supported by Projects MTM2016-78917-R and MTM2016-76969-P (AEI/FEDER, UE), grant AICO/2019/198 from Generalitat Valenciana, and IAP network StUDyS grant 3E120297 from the Belgian Science Policy

Gene expression and phenotype association can be affected by potential unmeasured confounders from multiple sources, leading to biased estimates of the associations. Since genetic variants largely explain gene expression variations, they can be used as instruments in studying the association between gene expressions and phenotype in the framework of high dimensional instrumental variable (IV) regression. However, because the dimensions of both genetic variants and gene expressions are often larger than the sample size, statistical inferences such as hypothesis testing for such high dimensional IV models are not trivial and have not been investigated in literature. The problem is more challenging since the instrumental variables (e.g., genetic variants) have to be selected among a large set of genetic variants. This paper considers the problem of hypothesis testing for sparse IV regression models and presents methods for testing single regression coefficient and multiple testing of multiple coefficients, where the test statistic for each single coefficient is constructed based on an inverse regression. A multiple testing procedure is developed for selecting variables and is shown to control the false discovery rate. Simulations are conducted to evaluate the performance of our proposed methods. These methods are illustrated by an analysis of a yeast dataset in order to identify genes that are associated with growth in the presence of hydrogen peroxide.

In this paper, we study a dual risk model with delays in the spirit of Dassios-Zhao. When a new innovation occurs, there is a delay before the innovation turns into a profit. We obtain large initial surplus asymptotics for the ruin probability and ruin time distributions. For some special cases, we get closed-form formulas. Numerical illustrations will also be provided., Comment: 17 pages, 2 figures, 2 tables

In this paper, we make use of meta-elliptical copula functions to build a new multivariate distribution with fixed marginal distributions and dependence structure to analyze bounded data. Specifica...

For a random sample of n iid p-dimensional vectors, each partitioned into b sub-vectors of dimensions pi, i=1,…,b, tests for zero correlation of sub-vectors are presented when pi ≫ n and the distribution need not be normal. The test statistics are composed of U-statistics based estimators of the Frobenius norm measuring the distance between the null and alternative hypotheses. Asymptotic distributions of the tests are provided for n,pi → ∞, with their finite-sample performance demonstrated through simulations. Some related tests are discussed. A real data application is also given.

Various methods of combining individual p-values into one p-value are widely used in many areas of statistical applications. We say that a combining method is valid for arbitrary dependence (VAD) if it does not require any assumption on the dependence structure of the p-values, whereas it is valid for some dependence (VSD) if it requires some specific, perhaps realistic but unjustifiable, dependence structures. The trade-off between validity and efficiency of these methods is studied via analyzing the choices of critical values under different dependence assumptions. We introduce the notions of independence-comonotonicity balance (IC-balance) and the price for validity. In particular, IC-balanced methods always produce an identical critical value for independent and perfectly positively dependent p-values, a specific type of insensitivity to a family of dependence assumptions. We show that, among two very general classes of merging methods commonly used in practice, the Cauchy combination method and the Simes method are the only IC-balanced ones. Simulation studies and a real data analysis are conducted to analyze the sizes and powers of various combining methods in the presence of weak and strong dependence., Comment: 40 pages, 5 figures and 6 tables

In some supervised learning settings, the practitioner might have additional information on the features used for prediction. We propose a new method which leverages this additional information for better prediction. The method, which we call the feature-weighted elastic net ("fwelnet"), uses these "features of features" to adapt the relative penalties on the feature coefficients in the elastic net penalty. In our simulations, fwelnet outperforms the lasso in terms of test mean squared error and usually gives an improvement in true positive rate or false positive rate for feature selection. We also apply this method to early prediction of preeclampsia, where fwelnet outperforms the lasso in terms of 10-fold cross-validated area under the curve (0.86 vs. 0.80). We also provide a connection between fwelnet and the group lasso and suggest how fwelnet might be used for multi-task learning.

Markov chain Monte Carlo (MCMC) is the engine of modern Bayesian statistics, being used to approximate the posterior and derived quantities of interest. Despite this, the issue of how the output from a Markov chain is post-processed and reported is often overlooked. Convergence diagnostics can be used to control bias via burn-in removal, but these do not account for (common) situations where a limited computational budget engenders a bias-variance trade-off. The aim of this article is to review state-of-the-art techniques for post-processing Markov chain output. Our review covers methods based on discrepancy minimisation, which directly address the bias-variance trade-off, as well as general-purpose control variate methods for approximating expected quantities of interest., Comment: Version 3 is the accepted version. When citing this paper, please use the following: South, LF, Riabiz, M, Teymur, O & Oates, CJ. 2022. Post-Processing of MCMC. Annual Review of Statistics and Its Application. 9: Submitted. DOI: 10.1146/annurev-statistics-040220-091727

Given a Wiener process with unknown and unobservable drift, we try to estimate this drift as effectively but also as quickly as possible, in the presence of a quadratic penalty for the estimation error and of a fixed, positive cost per unit of observation time. In a Bayesian framework, where the unobservable drift is assumed to have a known “prior” distribution, this question reduces to choosing judiciously a stopping time for an appropriate diffusion process in natural scale. We establish structural properties of the solution for the corresponding problem of optimal stopping. In particular, we show that, regardless of the prior distribution, the continuation region is monotonically shrinking in time. Moreover, we provide conditions on the prior distribution that guarantee a one-sided stopping region. Lastly, some concrete prior distributions are studied to illustrate the theoretical results.

This article generalises the concept of realised covariation to Hilbert-space-valued stochastic processes. More precisely, based on high-frequency functional data, we construct an estimator of the trace-class operator-valued integrated volatility process arising in general mild solutions of Hilbert space-valued stochastic evolution equations in the sense of Da Prato and Zabczyk (2014). We prove a weak law of large numbers for this estimator, where the convergence is uniform on compacts in probability with respect to the Hilbert-Schmidt norm. In addition, we show that the conditions on the volatility process are valid for most common stochastic volatility models in Hilbert spaces., Comment: 34 pages

Matrix-variate logistic regression is useful in facilitating the relationship between the binary response and complex-featured matrix-variates arising commonly from medical imaging research. However, standard inference procedures based on such a model are impaired by the presence of the response misclassification as well as inactive covariates. It is imperative to account for the misclassification effects and select active covariates when employing matrix-variate logistic regression to analyze such data. In this paper, we develop penalized unbiased estimating functions using the smoothly clipped absolute deviation (SCAD) penalty to address the sparsity of matrix-variate data as well as the response misclassification effects. The proposed methods are justified both theoretically and numerically. We analyze the Breast Cancer Wisconsin data with the proposed methods.

We construct a generalisation of Bakry-Emery curvature to prove twisted intertwining relations for Markov semigroups. These relations are applied to Brascamp–Lieb type inequalities and spectral gap results. It extends the method of Arnaudon, Bonnefont and Joulin, to Riemannian manifolds and to a wider class of twists. These results are illustrated with several examples.

In this article we extend the framework of explanatory mixed IRT models to a more general class called explanatory additive IRT models. We do this by augmenting the linear predictors in terms of smooth functions. This development offers many new modeling options such as the inclusion of nonlinear covariate effects, the specification of various temporal and spatial dependency patterns, and parameter partitioning across covariates. We use integrated nested Laplace approximation (INLA) for accurate and computationally efficient estimation of the parameters. Uninformative, weakly informative, and informative prior settings for the hyperparameters are discussed. Running time experiments and Monte Carlo parameter recovery simulations are performed in order to study the accuracy and computational efficiency of INLA when applied to the proposed explanatory additive IRT model class. Using a real-life dataset, a variety of application scenarios is explored, and the results are compared with classical maximum likelihood estimation when possible. R code is included in the supplemental materials to allow readers to fully reproduce the examples computed in the paper.

Techniques to reduce the energy burden of an industrial ecosystem often require solving a multiobjective optimization problem. However, collecting experimental data can often be either expensive or time-consuming. In such cases, statistical methods can be helpful. This article proposes Pareto Front (PF) and Pareto Set (PS) estimation methods using Bayesian Additive Regression Trees (BART), which is a non-parametric model whose assumptions are typically less restrictive than popular alternatives, such as Gaussian Processes (GPs). These less restrictive assumptions allow BART to handle scenarios (e.g. high-dimensional input spaces, nonsmooth responses, large datasets) that GPs find difficult. The performance of our BART-based method is compared to a GP-based method using analytic test functions, demonstrating convincing advantages. Finally, our BART-based methodology is applied to a motivating engineering problem. Supplementary materials, which include a theorem proof, algorithms, and R code, for this article are available online., Comment: 27 pages, 8 figures, submitted to Industry 4.0 special issue of Technometrics journal

This paper presents a Hybrid Feature Selection Technique for Sentiment Classification. We have used a Genetic Algorithm and a combination of existing Feature Selection methods, namely: Information Gain (IG), CHI Square (CHI), and GINI Index (GINI). First, we have obtained features from three different selection approaches as mentioned above and then performed the UNION SET Operation to extract the reduced feature set. Then, Genetic Algorithm is applied to optimize the feature set further. This paper also presents an Ensemble Approach based on the error rate obtained different domain datasets. To test our proposed Hybrid Feature Selection and Ensemble Classification approach, we have considered four Support Vector Machine (SVM) classifier variants. We have used UCI ML Datasets of three domains namely: IMDB Movie Review, Amazon Product Review and Yelp Restaurant Reviews. The experimental results show that our proposed approach performed best in all three domain datasets. Further, we also presented T-Test for Statistical Significance between classifiers and comparison is also done based on Precision, Recall, F1-Score, AUC and model execution time.

Semi-supervised clustering technique partitions the unlabeled data based on prior knowledge of labeled data. Most of the semi-supervised clustering algorithms exist only for the clustering of complete data, i.e., the data sets with no missing features. In this paper, an effort has been made to check the effectiveness of semi-supervised clustering when applied to incomplete data sets. The novelty of this approach is that it considers the missing features along with available knowledge (labels) of the data set. The linear interpolation imputation technique initially imputes the missing features of the data set, thus completing the data set. A semi-supervised clustering is now employed on this complete data set, and missing features are regularly updated within the clustering process. In the proposed work, the labeled percentage range used is 30, 40, 50, and 60% of the total data. Data is further altered by arbitrarily eliminating certain features of its components, which makes the data incomplete with partial labeling. The proposed algorithm utilizes both labeled and unlabeled data, along with certain missing values in the data. The proposed algorithm is evaluated using three performance indices, namely the misclassification rate, random index metric, and error rate. Despite the additional missing features, the proposed algorithm has been successfully implemented on real data sets and showed better/competing results than well-known standard semi-supervised clustering methods.

We present an exact distribution for the bounded range of a Wiener process by using the truncation method of the probability distributions. Rather than deriving some statistical properties of this ...

Summary The classical approach to analyze pharmacokinetic (PK) data in bioequivalence studies aiming to compare two different formulations is to perform noncompartmental analysis (NCA) followed by two one-sided tests (TOST). In this regard, the PK parameters area under the curve (AUC) and $C_{\max}$ are obtained for both treatment groups and their geometric mean ratios are considered. According to current guidelines by the U.S. Food and Drug Administration and the European Medicines Agency, the formulations are declared to be sufficiently similar if the $90\%$ confidence interval for these ratios falls between $0.8$ and $1.25 $. As NCA is not a reliable approach in case of sparse designs, a model-based alternative has already been proposed for the estimation of $\rm AUC$ and $C_{\max}$ using nonlinear mixed effects models. Here we propose another, more powerful test than the TOST and demonstrate its superiority through a simulation study both for NCA and model-based approaches. For products with high variability on PK parameters, this method appears to have closer type I errors to the conventionally accepted significance level of $0.05$, suggesting its potential use in situations where conventional bioequivalence analysis is not applicable.

We investigate the parameter estimation of regression models with fixed group effects, when the group variable is missing while group related variables are available. This problem involves clustering to infer the missing group variable based on the group related variables, and regression to build a model on the target variable given the group and eventually additional variables. Thus, this problem can be formulated as the joint distribution modeling of the target and of the group related variables. The usual parameter estimation strategy for this joint model is a two-step approach starting by learning the group variable (clustering step) and then plugging in its estimator for fitting the regression model (regression step). However, this approach is suboptimal (providing in particular biased regression estimates) since it does not make use of the target variable for clustering. Thus, we claim for a simultaneous estimation approach of both clustering and regression, in a semi-parametric framework. Numerical experiments illustrate the benefits of our proposition by considering wide ranges of distributions and regression models. The relevance of our new method is illustrated on real data dealing with problems associated with high blood pressure prevention.

Let denote $S_n(p) = k_n^{-1} \sum_{i=1}^{k_n} \left( \log (X_{n+1-i,n} / X_{n-k_n, n}) \right)^p$, where $p > 0$, $k_n \leq n$ is a sequence of integers such that $k_n \to \infty$ and $k_n / n \to 0$, and $X_{1,n} \leq \ldots \leq X_{n,n}$ is the order statistics of iid random variables with regularly varying upper tail. The estimator $\widehat \gamma(n) = (S_n(p)/\Gamma(p+1))^{1/p}$ is an extension of the Hill estimator. We investigate the asymptotic properties of $S_n(p)$ and $\widehat \gamma(n)$ both for fixed $p > 0$ and for $p = p_n \to \infty$. We prove strong consistency and asymptotic normality under appropriate assumptions. Applied to real data we find that for larger $p$ the estimator is less sensitive to the change in $k_n$ than the Hill estimator., Comment: 23 pages, 1 figure

Markov switching (MS-)GARCH(1,1) models allow for structural changes in volatility dynamics between a finite number of regimes. Since the regimes are not observed, computation of the likelihood requires integrating over an exponentially increasing number of regime paths, which is intractable. An existing smooth likelihood estimation procedure for sequential Monte Carlo (SMC), that is currently limited to hidden Markov models with a one-dimensional state variable, is modified to enable likelihood estimation and maximisation for MS-GARCH(1,1) models, a model which requires two dimensions, volatility and regime, to evolve its hidden state process. Furthermore, the modified SMC procedure is shown to be easily adapted to fitting MS-GARCH(1,1) models even when there are missing observations. The proposed methodology is validated with simulated data and is also illustrated with analysis of two financial time series, the daily returns on the S&P 500 index and on the Henry Hub natural gas spot price, with the latter series containing a gap caused by shutdown in response to hurricane Rita in 2005.

We study estimation of the correlation ρ in the multivariate linear model with uniform correlation structure. Unbiased estimator of ρ based on moment estimator proposed by Žežula (2006) is derived. Biased and unbiased estimators are compared by bias and mean square error under normal and t-distribution.

International audience; Being able to compare risk measures in practice is crucial in many applications such as in finance, insurance or environmental science. The difficulty is that the variables of interest are not always of the same nature, nor of the same type or scale. Thus the usual risk measures are often misleading and to solve this issue we propose to use the Expected Proportional Shortfall (EPS) which is scale invariant and thus, which does not depend on the unit of measurement. To estimate the EPS, an estimator of the tail index γ is required. The main asymptotic properties of our EPS estimator are provided under very general assumptions in case of d−variate β−mixing processes with Pareto-type marginals. Then, we propose a test statistic based on the EPS estimators to compare different risks, whatever their nature/type/scale are. Since the performances of the test statistic are poor when a biased estimate of γ is used, we propose to perform our EPS estimation with an asymptotically unbiased estimator for γ. The efficiency of our test statistic is illustrated in a simulation experiment and validated on an environmental dataset.