Change point analysis is concerned with detecting and locating structural breaks in the underlying model of a sequence of observations ordered by time, space or other variables. A widely adopted approach for change point analysis is to minimize an objective function with a penalty term on the number of change points. This framework includes several well-established procedures, such as the penalized log-likelihood using the (modified) Bayesian information criterion or the minimum description length. The resulting optimization problem can be solved in polynomial time by dynamic programming or its improved version, such as the pruned exact linear time (PELT) algorithm (Killick, Fearnhead, and Eckley 2012). However, existing computational methods often suffer from two primary limitations: (1) methods based on direct implementation of dynamic programming or PELT are often time-consuming for long data sequences due to repeated computation of the cost value over different segments of the data sequence; (2) state-ofthe-art R packages do not provide enough flexibility for users to handle different change point settings and models. In this work, we present the fastcpd package, aiming to provide an efficient and versatile framework for change point detection in several commonly encountered settings. The core of our algorithm is built upon PELT and the sequential gradient descent method recently proposed by Zhang and Dawn (2023). We illustrate the usage of the fastcpd package through several examples, including mean/variance changes in a (multivariate) Gaussian sequence, parameter changes in regression models, structural breaks in ARMA/GARCH/VAR models, and changes in user-specified models.
Lorenz regressions are statistical tools for measuring the extent of inequality in a response variable that is attributable to a set of covariates. In the context of a single-index model, these regression techniques estimate the explained Gini coefficient, a measure of inequality in the conditional expectation of the response given the covariates. In this paper, we describe the LorenzRegression package, which implements the non-penalized and penalized Lorenz regressions. The non-penalized procedure is implemented via a genetic algorithm, making use of the GA package. For the penalized case, the user can choose between using a lasso or SCAD penalty. In the former case, the estimation procedure is performed with the FABS algorithm, while the latter is implemented with the SCAD-FABS algorithm. Computationally intensive parts of the code are written in C++. A mathematical description of the procedure is provided. The main functions and S3 methods are carefully described and their use is illustrated on income data.
This paper introduces OGBoost, a scikit-learn-compatible Python package for ordinal regression gradient boosting. Ordinal variables (e.g., rating scales, quality assessments) lie between nominal and continuous data, requiring specialized methods that respect their inherent ordering. OGBoost employs a novel coordinate-descent optimization approach within the cumulative link framework, jointly optimizing a continuous regression function via functional gradient descent and a threshold vector via standard gradient descent. Key features include: (i) latent score predictions via decision_function for enhanced ranking performance; (ii) heterogeneous boosting ensembles by enabling any combination of scikit-learn regressors as base learners; and (iii) cross-validation-based early stopping for improved robustness. We provide a comprehensive review of existing ordinal regression software and present empirical comparisons across 17 public datasets, demonstrating that OGBoost significantly outperforms representative methods including OrdinalGBT, scikitlearn's gradient boosting classifier, and scikit-lego's ordinal classifier. Formal statistical tests confirm OGBoost's superiority on concordance index (achieving best performance on all 17 datasets), with consistent performance advantages across classification accuracy, mean absolute error, and Spearman correlation. The package integrates seamlessly with scikit-learn machine learning workflows and is available via pip install ogboost.
This study presents a fast and consistent software package called BACE (i.e., Bayesian averaging of classical estimates) that offers a model-building strategy for various bounded dependent variable models, including-logit and probit models, ordered logit and pro-bit models, multinomial logistic regression, Poisson regression, Tobit model, interval regression-as well as linear regression. The BACE approach, originally associated with normal linear regression, is a model selection method that combines classical estimation and Bayesian techniques. It solves the problem of computational speed and model uncertainty that arise when dealing with numerous competing advanced statistical models. Our package also provides an implementation of the well-established Bayesian information criterion variants and the latest measures of jointness. We use gretl, a popular, free, and open-source software for econometric analysis that features an easy-to-use graphical user interface.
asgl is an open-source Python package that offers a robust and versatile framework for fitting a variety of regression models including linear, logistic, and, notably, quantile regression. It implements a comprehensive suite of penalization techniques such as LASSO, ridge, group LASSO, sparse group LASSO, elastic net, and their adaptive variants. A key contribution of asgl is its extensive support for adaptive penalizations, critically offering a range of built-in methodologies for estimating the necessary adaptive weights as proposed by Mendez-Civieta, Aguilera-Morillo, and Lillo (2021). This feature addresses a significant practical challenge-the weight estimation process-in applying advanced adaptive methods, especially in high-dimensional settings, and is largely absent from other packages. Furthermore, asgl offers penalized quantile regression, a less commonly available feature in statistical software. The primary class, Regressor, ensures seamless integration with the scikit-learn ecosystem, facilitating straightforward model evaluation and hyperparameter optimization. asgl has demonstrated utility in variable selection and prediction tasks across both low-and high-dimensional data, positioning it as a comprehensive tool for modern statistical modeling.
