The March issue of the CJS comes under the sign of spring renewal. On the one hand, it is the first issue which is merely electronic. On the other, a more stringent data sharing policy has been implemented to ensure better reproducibility of the data analyses. Authors are now expected to share their data unless there are privacy or ethical concerns, and they are also strongly encouraged to share their code. You will see this new editorial policy take effect in papers that will start appearing in early view shortly, once Wiley has ironed out a few remaining wrinkles of their emulation of our new design.

This issue is already available online and consists of 14 research articles. It opens with two contributions to biostatistics. Zhang, Yin, and Rubin [1] propose a novel rerandomization strategy to improve estimation of average treatment effects in high-dimensional settings. Prior to calculating covariate means using the Mahalanobis distance, their idea is to first identify suitable covariate subspaces using principal component analysis, thereby significantly reducing dimensionality and increasing computational simplicity. For their part, Le, Bai, and Qin [2] suggest a simultaneous strategy for subgroup analysis assuming a linear regression model in order to identify different population subgroups in which a given treatment has different effects and to estimate these effects when measurement errors are present.

The next two papers consider testing problems. To detect differences between covariance operators of several populations of functional data, Ramsay and Chenouri [3] propose robust, nonparametric Kruskal-Wallis tests which rely on functional data depth ranks. Zhao and Sun [4] develop tests to identify polygenetic signals in high-dimensional generalized linear models for genetic association studies of complex traits. Their procedure utilizes repeated sample splitting to ensure valid and stable post-variable-selection inference.

The next five articles contribute to regression modeling. Ordoñez, Prates, Bazán, and Lachos [5] introduce a penalized complexity prior for the skewness parameter of power links in generalized linear models for binary, binomial or bounded data. This skewness parameter provides extra modeling flexibility, e.g., when dealing with imbalanced data. As for Basa, Cook, Forzani, and Marcos [6], they establish the asymptotic properties of the partial least squares estimator in one-component linear regression models when the number of covariates grows with the sample size. To help identify subgroups in complex, heterogeneous populations, Liu and Li [7] propose a model averaging approach for segment regression models with multiple threshold variables and multiple structural breaks. Wen, Chen, Wang, and Pan, for the Alzheimer’s Disease Neuroimaging Initiative [8] consider variable selection in generalized additive models. Their novel penalized procedure, which enjoys support recovery consistency, allows them to perform joint selection of variables and basis functions. Tsao [9] contributes a novel method for model selection for general regression models based on likelihood ratio tests.

The paper by Feng, Tang, and Ding [10] advances inference for incomplete data, an omnipresent issue in applications. They develop both graphical model validation techniques and formal goodness-of-fit tests for additive hazards models for case II interval-censored data.

The article by Yuan, Zhou, Zhang, and Cui [11] contributes to financial data modeling. To forecast volatility of financial securities, a GARCH-Itô-type model is used that integrates three major information sources: low and high frequency historical price data and option data. Instead of using option prices directly, however, their procedure rests on option-implied volatility.

The next pair of articles offers solutions for selected issues that arise when handing massive data. On one hand, Chen, Wang, and Chang [12] show how to conduct several adaptive sequential procedures simultaneously in a way that ensures similar statistical properties of the resulting estimates. This facilitates their subsequent integration, as the authors detail on two-sided confidence set estimation in a linear model. On the other hand, Wang, Wang, and Xiong [13] consider optimal subsampling schemes under measurement constraints that are particularly convenient when responses are expensive to measure. They propose an unweighted estimator and show, using martingale methods, that it is more efficient than existing weighted techniques.

The closing article by Li and Sun [14] contributes a novel way of constructing space-filling designs for computer experiments. Their proposal has more economical run sizes, fulfills column-orthogonality, and enjoys many desirable low-dimensional stratification properties, in addition to being able to generate designs with 3-orthogonality.  

Wishing you inspirational readings,

Johanna G. Nešlehová
Editor-in-Chief, The Canadian Journal of Statistics

Table of Contents of the March 2024 Issue of The Canadian Journal of Statistics

[1]    PCA rerandomization, by/par Hengtao Zhang, Guosheng Yin & Donald B. Rubin
[2]    Subgroup analysis of linear models with measurement error, by/par Yuan Le, Yang Bai & Guoyou Qin
[3]    Robust nonparametric hypothesis tests for differences in the covariance structure of functional data, by/par Kelly Ramsay & Shoja’eddin Chenouri
[4]    A stable and adaptive polygenic signal detection method based on repeated sample splitting, by/par Yanyan Zhao & Lei Sun
[5]    Penalized complexity priors for the skewness parameter of power links, by/par José A. Ordoñez, Marcos O. Prates, Jorge L. Bazán & Victor H. Lachos
[6]    Asymptotic distribution of one-component partial least squares regression estimators in high dimensions, by/par Jerónimo Basa, R. Dennis Cook, Liliana Forzani & Miguel Marcos
[7]    Segment regression model average with multiple threshold variables and multiple structural breaks, by/par Pan Liu & Jialiang Li
[8]    Variable selection in additive models via hierarchical sparse penalty, by/par Canhong Wen, Anan Chen, Xueqin Wang & Wenliang Pan, for the Alzheimer’s Disease Neuroimaging Initiative
[9]    Regression model selection via log-likelihood ratio and constrained minimum criterion, by/par Min Tsao
[10]    Method of model checking for case II interval-censored data under the additive hazards model, by/par Yanqin Feng, Ming Tang & Jieli Ding
[11]    Volatility analysis for the GARCH-Itô model with option data, by/par Huiling Yuan, Yong Zhou, Zhiyuan Zhang & Xiangyu Cui
[12]    Distributed sequential estimation procedures, by/par Zhuojian Chen, Zhanfeng Wang & Yuan-chin Ivan Chang
[13]    Unweighted estimation based on optimal sample under measurement constraints, by/par Jing Wang, HaiYing Wang & Shifeng Xiong
[14]    A class of space-filling designs with low-dimensional stratification and column orthogonality, by/par Pengnan Li & Fasheng Sun