While listening to Christmas carols as I write this editorial, I feel nostalgic about the days when one would take the time to leaf through new editions of journals in the library or a departmental common room. If you were to open a paper version of the December issue of CJS today, you would immediately notice that it looks quite different because this issue is the first to have been produced with the new LaTeX class designed by Vincent Goulet. Yey! You would also note that the opening article is at the same time the inaugural contribution to the CJS celebration issue, honouring award-winning Canadian statisticians, junior and senior. It features Ruodu Wang, from the University of Waterloo. More articles of this type will come and will be highlighted at the following website, along with biosketches and prize citations:

https://onlinelibrary.wiley.com/doi/toc/10.1002/(ISSN)1708-945X.celebra…

Never mind nostalgia, printed journal issues are a thing of the past. CJS is already available only in electronic format, but the December issue is the first in which traditional page numbers have been replaced by e-locators.

I am glad to be able to introduce these innovations today, because this issue is the last that I will put together in my capacity as CJS editor. It has been an honour serving you in this role, and an enriching and gratifying experience to shepherd the publication process of so many original contributions. My sincere thanks are due to my amazing team of associate editors, but also my editorial assistant, Julie Falkner, who is immensely devoted, as well as the ever so supportive and professional CJS managing editor and SSC publications committee chair, Llwellyn Armstrong and Rhonda Rosychuk.

One of the trademarks of the CJS is its support of French as a language of scientific communication. Llwellyn and I have worked together with Wiley to make the journal’s website bilingual; likewise, the LaTeX class cjs-rcs-article and its documentation now also fully support Canada’s two official languages. I am grateful to Belkacem Abdous for the translation of abstracts, and much indebted to my husband, Christian Genest, for translating and proofreading numerous texts, including all my editorials for Liaison.

The journal wouldn’t be what it is without the devotion and support of its authors, reviewers, copy editors, and readers—sincere thanks to you all! You have much more to look forward to during the term of the incoming editor in chief, Alexandra Schmidt, who like me is affiliated to McGill University. She brings extensive experience from which the journal will surely greatly benefit. I wish her all possible success in her new role.

Without further ado, let me now give you a digest of the content of the December issue.

The first article by Vovk and Wang [1] will plunge you in the world of e-values in the context of multiple hypothesis testing. New algorithms are developed for the case in which the base tests produce independent or sequential e-values. Hypothesis testing is also the topic of the contribution by Zhuang, Wang, and Chen [2], who aim to test for first-order stochastic dominance of two distributions based on an independent sample from each. Their tests are particularly sensitive to violation of the stochastic dominance relationship in the tails of the distributions. Zhang, Liu, and Wu [3] consider one-sided hypothesis tests for nonlinear mixed-effects models for longitudinal data. An advantage is a more realistic setup, which allows for measurement errors in the covariates.

Wang [4] considers a regression model in which the response is some function of the covariates. As the form of this curve may differ for some batches of responses, a nonparametric procedure to estimate the location and the number of such changepoints is introduced. De Silva and Choudhary [5] model functional binomial or Poisson responses via generalized functional principal component analysis, and construct pointwise and simultaneous tolerance bands for them.

Torkashvand and Jafari Jozani [6] study small-area estimation of parameters associated with positive quantities such as income. They introduce the so-called weighted precautionary loss function and obtain constrained hierarchical Bayes estimates that are better at avoiding underestimation of, e.g., disease rates. Nguyen and Jiang [7] estimate the mean squared prediction error for the observed best prediction in small-area estimation with count data. Their idea is to utilize the Prasad-Rao-type linearization method but adapt it to a setting in which the model is possibly misspecified.

Frydman and Surya [8] develop maximum likelihood estimation for a general mixture of Markov jump processes, whose generator matrices are unconstrained. They then use it as an exploratory tool to identify homogeneous subpopulations of a heterogeneous population.

Occasionally, it can happen that a model was built from a source data with a different covariate distribution than the target population. Assuming that covariate data from the target population is at hand, Morrison, Gatsonis, Dahabreh, Li, and Steingrimsson [9] propose robust estimation of loss-based measures of model performance under such “covariate shift.”

The next two articles pertain to interval-censored data. Ma, Wang, Lou, and Sun [10] consider data from case-cohort studies that are subject to informative interval-censoring and are such that the covariates are collected only for a small subcohort. They use an additive hazard model and propose a sieve inverse probability weighting procedure for its estimation. Yang, Li, Diao, and Cook [11] develop predictive algorithms when training samples involve interval censoring. The delicate issue here is to define an observed data loss function in a way that it is an unbiased estimate of the loss function based on the unobservable complete data.

The issue closes with a statistical lesson learned from COVID-19. Wu, Stephens, and Moodie [12] study the Susceptible-Infectious-Recovered epidemic model when only the time series of deaths and seroprevalence survey data are available. They develop a Bayesian framework for infection estimation and test its robustness to the more realistic but also more complex Susceptible-Exposed-Infectious-Recovered-based epidemics.

Wishing you inspirational readings as well as happy and healthy holidays!

Johanna G. Nešlehová, Editor-in-Chief

The Canadian Journal of Statistics

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

1. True and false discoveries with independent and sequential e-values, by/par Vladimir Vovk & Ruodu Wang

2. Tests for the first-order stochastic dominance, by/par Weiwei Zhuang, Peiming Wang, & Jiahua Chen 

3. Order-restricted hypothesis tests for nonlinear mixed-effects models with measurement errors in covariates, by/par Yixin Zhang, Wei Liu, & Lang Wu

4. Multiple change-point detection for regression curves, by/par Yunlong Wang

5. Tolerance bands for exponential family functional data, by/par Galappaththige S. R. de Silva & Pankaj K. Choudhary

6. Constrained Bayes in multiplicative area-level models under the precautionary loss function, by/par Elaheh Torkashvand & Mohammad Jafari Jozani

7. Estimating the mean squared prediction error of the observed best predictor associated with small area counts: A computationally oriented approach, by/par Thuan Nguyen & Jiming Jiang

8. Estimation in a general mixture of Markov jump processes, by/par Halina Frydman & Budhi Arta Surya

9. Robust estimation of loss-based measures of model performance under covariate shift, by/par Samantha Morrison, Constantine Gatsonis, Issa J. Dahabreh, Bing Li, & Jon A. Steingrimsson

10. Estimation of the additive hazards model based on case-cohort interval-censored data with dependent censoring, by/par Yuqing Ma, Peijie Wang, Yichen Lou, & Jianguo Sun

11. Regression trees for interval-censored failure time data based on censoring unbiased transformations and pseudo-observations, by/par Ce Yang, Xianwei Li, Liqun Diao, & Richard J. Cook

12. An SIR-based Bayesian framework for COVID-19 infection estimation, by/par Haoyu Wu, David A. Stephens, & Erica E. M. Moodie