Dan Jeske, The American Statistician Editor

The November 2019 issue of The American Statistician features 11 articles and two letters to the editor.

The General section has three articles. The first looks into tests for trends over time in multinomial probabilities. A test of no trend versus the alternative of a trend for at least one of the categories is discussed. The second article formulates a two-sided test by decomposing the alternative into multiple directional alternatives. The test procedure is a set of rules that maps the rejection decision to a specific alternative. The third article addresses the replication crisis through a Bayesian lens. A Bayesian replication factor developed for a two-group setting is extended to a more general multi-group setting.

The Statistical Practice section has two articles. The first is concerned with missing data issues in logistic regression. Nonignorable missing responses are considered and a method to reduce the bias in the estimated regression coefficients is presented. The second paper delves into power and sample size considerations for linear mixed models. Mapping the linear mixed model to a multivariate linear model is key to making these calculations more tractable.

There are two articles in the Teacher’s Corner. The first provides powerful geometric intuition for the tail formula used to compute the expected value of a random variable. Several interesting examples are given, and multivariate extensions are discussed. The second paper is a comparative review of alternative textbooks that might be used for an undergraduate course in nonparametric statistics.

An article in the Statistical Computing and Graphics section proposes a framework for assessing statistical computing tools. The framework could be useful for developers of statistical tools and/or educators using statistical tools in their courses.

An article in the Interdisciplinary section uses spatial generalized linear models to study the allocation of foreign aid to Malawi. The purpose of the modeling effort is to understand whether the aid is being delivered efficiently to the areas that most need it.

An article in the History Corner reviews the history of industrial statistics and profiles several pioneering industrial statistics organizations.

A short technical note in this issue revisits the mean value theorem for vector-valued estimating functions. The article is a follow-up to a previously published TAS paper on this topic.

The November issue concludes with two letters to the editor. The first addresses a TAS paper that modeled baseball slugging percentage and the second raises an interesting question about how power calculations should proceed if a significance threshold cutoff (e.g., .05) is to be avoided.

To view the articles, visit Taylor & Francis online or log in at www.amstat.org.