Asta-Advances in Statistical Analysis最新文献

A pandemic poses particular challenges to decision-making because of the need to continuously adapt decisions to rapidly changing evidence and available data. For example, which countermeasures are appropriate at a particular stage of the pandemic? How can the severity of the pandemic be measured? What is the effect of vaccination in the population and which groups should be vaccinated first? The process of decision-making starts with data collection and modeling and continues to the dissemination of results and the subsequent decisions taken. The goal of this paper is to give an overview of this process and to provide recommendations for the different steps from a statistical perspective. In particular, we discuss a range of modeling techniques including mathematical, statistical and decision-analytic models along with their applications in the COVID-19 context. With this overview, we aim to foster the understanding of the goals of these modeling approaches and the specific data requirements that are essential for the interpretation of results and for successful interdisciplinary collaborations. A special focus is on the role played by data in these different models, and we incorporate into the discussion the importance of statistical literacy and of effective dissemination and communication of findings.

In this paper, we consider the confidence interval construction for the partially nonlinear models with missing responses at random under the framework of quantile regression. We propose an imputation-based empirical likelihood method to construct statistical inferences for both the unknown parametric vector in the nonlinear function and the nonparametric function and show that the proposed empirical log-likelihood ratios are both asymptotically chi-squared in theory. Furthermore, the confidence region for the parametric vector and the pointwise confidence interval for the nonparametric function are constructed. Some simulation studies are implemented to assess the performances of the proposed estimation method, and simulation results indicate that the proposed method is workable.

The Riesz probability distribution on symmetric matrices represents an important extension of the Wishart distribution. It is defined by its Laplace transform involving the notion of generalized power. Based on the fact that some Wishart distributions are presented by the mean of the multivariate Gaussian distribution, it is shown that some Riesz probability distributions which are not necessarily Wishart are also presented by the mean of Gaussian samples with missing data. As a corollary, we deduce a Gaussian representation of the inverse Riesz distribution and we give its expectation. The results are assessed in simulation studies.

Sustainability of agriculture is difficult to measure and assess because it is a multidimensional concept that involves economic, social and environmental aspects and is subjected to temporal evolution and geographical differences. Existing studies assessing agricultural sustainability in the European Union (EU) are affected by several shortcomings that limit their relevance for policy makers. Specifically, most of them focus on farm level or cover a small set of countries, and the few exceptions covering a broad set of countries consider only a subset of the sustainable dimensions or rely on cross-sectional data. In this paper, we consider yearly data on 12 indicators (5 for the economic, 3 for the social and 4 for the environmental dimension) measured on 26 EU countries in the period 2004–2018 (15 years), and apply group-based multivariate trajectory modeling to identify groups of countries with common trends of sustainable objectives. An expectation-maximization algorithm is proposed to perform maximum likelihood estimation from incomplete data without relying on an explicit imputation procedure. Our results highlight three groups of countries with distinguished strong and weak sustainable objectives. Strong objectives common to all the three groups include improvement of productivity, increase of personal income in rural areas, reduction of poverty in rural areas, increase of production of renewable energy, rise of organic farming and reduction of nitrogen balance. Instead, enhancement of manager turnover and reduction of greenhouse gas emissions are weak objectives common to all the three groups of countries. Our findings represent a valuable resource to formulate new schemes for the attribution of subsidies within the Common Agricultural Policy (CAP).

We present a data-driven approach to predict the next action in soccer. We focus on passing actions of the ball possessing player and aim to forecast the pass itself and when, in time, the pass will be played. At the same time, our model estimates the probability that the player loses possession of the ball before she can perform the action. Our approach consists of parameterized exponential rate models for all possible actions that are adapted to historic data with graph recurrent neural networks to account for inter-dependencies of the output space (i.e., the possible actions). We report on empirical results.

Prediction of quantiles at extreme tails is of interest in numerous applications. Extreme value modelling provides various competing predictors for this point prediction problem. A common method of assessment of a set of competing predictors is to evaluate their predictive performance in a given situation. However, due to the extreme nature of this inference problem, it can be possible that the predicted quantiles are not seen in the historical records, particularly when the sample size is small. This situation poses a problem to the validation of the prediction with its realization. In this article, we propose two non-parametric scoring approaches to assess extreme quantile prediction mechanisms. The proposed assessment methods are based on predicting a sequence of equally extreme quantiles on different parts of the data. We then use the quantile scoring function to evaluate the competing predictors. The performance of the scoring methods is compared with the conventional scoring method and the superiority of the former methods are demonstrated in a simulation study. The methods are then applied to analyze cyber Netflow data from Los Alamos National Laboratory and daily precipitation data at a station in California available from Global Historical Climatology Network.

Count data as response variables are commonly modeled using Poisson regression models, which require equidispersion, i.e., equal mean and variance. However, this relationship does not always occur, and the variance may be higher or lower than the mean, phenomena are known as overdispersion and underdispersion, respectively. Non-equidispersion, when disregarded, can lead to a number of misinterpretations and inadequate predictions. Here, we compare the use of the COM-Poisson, double Poisson, Gamma-count, and restricted generalized Poisson models as a more flexible class for count problems associated with over- and underdispersion, since they have an additional parameter that allows more flexible analysis. The proposed method is useful in different applications, but here we provide an example using an underdispersed dataset concerning ecological invasion. For validation of the models, we use half-normal plots. The COM-Poisson, double Poisson, and Gamma-count performed best and properly modeled the underdispersion. The use of correct statistical models is recommended to handle this data property using objective criteria to ensure accurate statistical inferences.

We explore and illustrate the concept of ranked sparsity, a phenomenon that often occurs naturally in modeling applications when an expected disparity exists in the quality of information between different feature sets. Its presence can cause traditional and modern model selection methods to fail because such procedures commonly presume that each potential parameter is equally worthy of entering into the final model—we call this presumption “covariate equipoise.” However, this presumption does not always hold, especially in the presence of derived variables. For instance, when all possible interactions are considered as candidate predictors, the premise of covariate equipoise will often produce over-specified and opaque models. The sheer number of additional candidate variables grossly inflates the number of false discoveries in the interactions, resulting in unnecessarily complex and difficult-to-interpret models with many (truly spurious) interactions. We suggest a modeling strategy that requires a stronger level of evidence in order to allow certain variables (e.g., interactions) to be selected in the final model. This ranked sparsity paradigm can be implemented with the sparsity-ranked lasso (SRL). We compare the performance of SRL relative to competing methods in a series of simulation studies, showing that the SRL is a very attractive method because it is fast and accurate and produces more transparent models (with fewer false interactions). We illustrate its utility in an application to predict the survival of lung cancer patients using a set of gene expression measurements and clinical covariates, searching in particular for gene–environment interactions.