Computational Statistics最新文献

In a context of component-based multivariate modeling we propose to model the residual dependence of the responses. Each response of a response vector is assumed to depend, through a Generalized Linear Model, on a set of explanatory variables. The vast majority of explanatory variables are partitioned into conceptually homogeneous variable groups, viewed as explanatory themes. Variables in themes are supposed many and some of them are highly correlated or even collinear. Thus, generalized linear regression demands dimension reduction and regularization with respect to each theme. Besides them, we consider a small set of “additional” covariates not conceptually linked to the themes, and demanding no regularization. Supervised Component Generalized Linear Regression proposed to both regularize and reduce the dimension of the explanatory space by searching each theme for an appropriate number of orthogonal components, which both contribute to predict the responses and capture relevant structural information in themes. In this paper, we introduce random latent variables (a.k.a. factors) so as to model the covariance matrix of the linear predictors of the responses conditional on the components. To estimate the model, we present an algorithm combining supervised component-based model estimation with factor model estimation. This methodology is tested on simulated data and then applied to an agricultural ecology dataset.

Statistical learning methods are widely utilised in tackling complex problems due to their flexibility, good predictive performance and ability to capture complex relationships among variables. Additionally, recently developed automatic workflows have provided a standardised approach for implementing statistical learning methods across various applications. However, these tools highlight one of the main drawbacks of statistical learning: the lack of interpretability of the results. In the past few years, a large amount of research has been focused on methods for interpreting black box models. Having interpretable statistical learning methods is necessary for obtaining a deeper understanding of these models. Specifically in problems in which spatial information is relevant, combining interpretable methods with spatial data can help to provide a better understanding of the problem and an improved interpretation of the results. This paper is focused on the individual conditional expectation plot (ICE-plot), a model-agnostic method for interpreting statistical learning models and combining them with spatial information. An ICE-plot extension is proposed in which spatial information is used as a restriction to define spatial ICE (SpICE) curves. Spatial ICE curves are estimated using real data in the context of an economic problem concerning property valuation in Montevideo, Uruguay. Understanding the key factors that influence property valuation is essential for decision-making, and spatial data play a relevant role in this regard.

This paper investigates the effectiveness of various metrics for selecting the adequate model for binary classification when data is imbalanced. Through an extensive simulation study involving 12 commonly used metrics of classification, our findings indicate that the Matthews Correlation Coefficient, G-Mean, and Cohen’s kappa consistently yield favorable performance. Conversely, the area under the curve and Accuracy metrics demonstrate poor performance across all studied scenarios, while other seven metrics exhibit varying degrees of effectiveness in specific scenarios. Furthermore, we discuss a practical application in the financial area, which confirms the robust performance of these metrics in facilitating model selection among alternative link functions.

This paper presents a theory of contrasts designed for modified Freeman–Tukey (FT) statistics which are derived through square-root transformations of observed frequencies (proportions) in contingency tables. Some modifications of the original FT statistic are necessary to allow for ANOVA-like exact decompositions of the global goodness of fit (GOF) measures. The square-root transformations have an important effect of stabilizing (equalizing) variances. The theory is then used to derive Tukey’s post-hoc pairwise comparison tests for contingency tables. Tukey’s tests are more restrictive, but are more powerful, than Scheffè’s post-hoc tests developed earlier for the analysis of contingency tables. Throughout this paper, numerical examples are given to illustrate the theory. Modified FT statistics, like other similar statistics for contingency tables, are based on a large-sample rationale. Small Monte-Carlo studies are conducted to investigate asymptotic (and non-asymptotic) behaviors of the proposed statistics.

Nonconvex penalties are utilized for regularization in high-dimensional statistical learning algorithms primarily because they yield unbiased or nearly unbiased estimators for the parameters in the model. Nonconvex penalties existing in the literature such as SCAD, MCP, Laplace and arctan have a singularity at origin which makes them useful also for variable selection. However, in several high-dimensional frameworks such as deep learning, variable selection is less of a concern. In this paper, we present a nonconvex penalty which is smooth at origin. The paper includes asymptotic results for ordinary least squares estimators regularized with the new penalty function, showing asymptotic bias that vanishes exponentially fast. We also conducted simulations to better understand the finite sample properties and conducted an empirical study employing deep neural network architecture on three datasets and convolutional neural network on four datasets. The empirical study based on artificial neural networks showed better performance for the new regularization approach in five out of the seven datasets.

The power of data and correct statistical analysis has never been more prevalent. Academics and practitioners require nowadays an accurate application of quantitative methods. Yet many branches are subject to a crisis of integrity, which is shown in an improper use of statistical models, p-hacking, HARKing, or failure to replicate results. We propose the use of a Peer-to-Peer (P2P) ecosystem based on a blockchain network, Quantinar, to support quantitative analytics knowledge paired with code in the form of Quantlets or software snippets. The integration of blockchain technology allows Quantinar to ensure fully transparent and reproducible scientific research.

We make Bayesian additive regression networks (BARN) available as a Python package, barmpy, with documentation at https://dvbuntu.github.io/barmpy/ for general machine learning practitioners. Our object-oriented design is compatible with SciKit-Learn, allowing usage of their tools like cross-validation. To ease learning to use barmpy, we produce a companion tutorial that expands on reference information in the documentation. Any interested user can pip install barmpy from the official PyPi repository. barmpy also serves as a baseline Python library for generic Bayesian additive regression models.

Meta-analysis is an important statistical technique for synthesizing the results of multiple studies regarding the same or closely related research question. So-called meta-regression extends meta-analysis models by accounting for study-level covariates. Mixed-effects meta-regression models provide a powerful tool for evidence synthesis, by appropriately accounting for between-study heterogeneity. In fact, modelling the study effect in terms of random effects and moderators not only allows to examine the impact of the moderators, but often leads to more accurate estimates of the involved parameters. Nevertheless, due to the often small number of studies on a specific research topic, interactions are often neglected in meta-regression. In this work we consider the research questions (i) how moderator interactions influence inference in mixed-effects meta-regression models and (ii) whether some inference methods are more reliable than others. Here we review robust methods for confidence intervals in meta-regression models including interaction effects. These methods are based on the application of robust sandwich estimators of Hartung-Knapp-Sidik-Jonkman (HKSJ) or heteroscedasticity-consistent (HC)-type for estimating the variance-covariance matrix of the vector of model coefficients. Furthermore, we compare different versions of these robust estimators in an extensive simulation study. We thereby investigate coverage and width of seven different confidence intervals under varying conditions. Our simulation study shows that the coverage rates as well as the interval widths of the parameter estimates are only slightly affected by adjustment of the parameters. It also turned out that using the Satterthwaite approximation for the degrees of freedom seems to be advantageous for accurate coverage rates. In addition, different to previous analyses for simpler models, the (textbf{HKSJ})-estimator shows a worse performance in this more complex setting compared to some of the (textbf{HC})-estimators.

This work focuses on learning causal network structures from ordinal categorical data. By combining constraint-based with score-and-search methodologies in structural learning, we propose a hybrid method called Markov Blanket Based Ordinal Causal Discovery (MBOCD) algorithm, which can capture the ordinal relationship of values in ordinal categorical variables. Theoretically, it is proved that for ordinal causal networks, two adjacent DAGs belonging to the same Markov equivalence class are identifiable, which results in the generation of a causal graph. Simulation experiments demonstrate that the proposed algorithm outperforms existing methods in terms of computational efficiency and accuracy. The code of this work is open at: https://github.com/leoydu/MBOCDcode.git.

This paper introduces the Metropolis–Hastings variational inference Robbins–Monro (MHVIRM) algorithm, a modification of the Metropolis–Hastings Robbins–Monro (MHRM) method, designed for estimating parameters in complex multidimensional graded response models (MGRM). By integrating a black-box variational inference (BBVI) approach, MHVIRM enhances computational efficiency and estimation accuracy, particularly for models with high-dimensional data and complex test structures. The algorithms effectiveness is demonstrated through simulations, showing improved precision over traditional MHRM, especially in scenarios with complex structures and small sample sizes. Moreover, MHVIRM is robust to initial values. The applicability is further illustrated with a real dataset analysis.