Editorial

“Editorial” published in The R Journal.

In this issue

On behalf of the Editorial Board, I am pleased to present Volume 18, Issue 3 of the R Journal. This issue features twenty research articles spanning a wide range of developments in statistical methodology, machine learning, time series, spatial analysis, causal inference, visualisation, and scientific computing.

Several articles introduce new tools for computationally challenging statistical problems. hdtg provides methods for simulating from high-dimensional truncated multivariate normal distributions, while ODRF develops an implementation of oblique decision trees and random forests. psvmSDR provides tools for sufficient dimension reduction using principal machines, and DRIP brings jump regression methods to image analysis. CRANpkg{SelectionBias} implements a range of methods for bounding selection bias in causal estimands. LBBNN provides a scalable approach to sparse and interpretable Bayesian neural networks, making use of variational inference and GPU acceleration. The issue also includes rSRD supporting the sum of ranking differences procedure, and xplainfi brings together a broad range of feature importance methods with tools for statistical inference.

Time series modelling is also well represented in this issue. mtarm provides Bayesian estimation and forecasting for multivariate threshold autoregressive models, while echos introduces automated time series forecasting using echo state networks. MSTest provides a collection of procedures for testing the number of regimes in Markov-switching models, VARshrink brings together shrinkage methods for high-dimensional vector autoregressive models, and iAR provides tools for modelling irregularly observed time series. demofit brings parametric and stochastic approaches to mortality modelling and forecasting into a unified framework.

There are also several contributions in spatial, spatiotemporal, and multivariate analysis. ldmppr provides tools for modelling marked point processes where the marks may depend on location, while sugarglider extends glyph-map visualisation for exploring spatiotemporal data. goSorensen provides methods for comparing gene lists using Gene Ontology information, admix supports estimation, testing and clustering for admixture models, and dbrobust provides tools for robust distance-based analysis and visualisation of mixed-type data.

Finally, metafrontier provides a comprehensive set of tools for metafrontier analysis of efficiency and productivity.

All packages discussed are available on CRAN. Supplementary material with fully reproducible code is available for download from the R Journal website.

0.1 CRAN packages used

hdtg, ODRF, psvmSDR, DRIP, LBBNN, rSRD, xplainfi, mtarm, echos, MSTest, VARshrink, iAR, demofit, ldmppr, sugarglider, goSorensen, admix, dbrobust, metafrontier

0.2 CRAN Task Views implied by cited packages

Distributions, TimeSeries