The November 2018 issue of The American Statistician features 13 articles and one letter to the editor that span a range of interesting methodology and application areas. In keeping with the mission of TAS, there is something for everyone in this issue.
The General section begins with an article that develops Bayesian inference for Kendall’s rank correlation coefficient. It is interesting how this can be done in the absence of an explicit likelihood function.
A second article investigates alternative ways to linearly transform a vector of random variables into uncorrelated random variables. We might have in mind a way to do it, but odds are many of us have not thought about the alternative ways to do it that are described in this paper.
The next paper in this section presents surprising inference contexts where it is better to use the raw data than it is to use the sufficient statistics.
A fourth paper reports on a survey regarding the practices of authors who use simulation studies in their research and highlights what should be improved.
The General section concludes with a paper that introduces an alternative measure of (income) inequality and compares it to the well-known Gini coefficient.
That Statistical Practice section includes two papers. The first advocates for the use of shift alternatives when testing hypotheses and considers them in the context of using Wilcoxon’s signed-rank test. The second suggests modifications to boxplots that improve their applications to skewed data.
We have three articles in the Teacher’s Corner. The first proposes specific ways to teach critical thinking skills in an introductory statistics class, with the goal being to improve statistical literacy and prepare students to be better decision-makers.
The second article outlines an approach to teaching ethics in the context of a graduate-level statistics class and includes specific connections to the ethical guidelines prepared by the ASA.
The final article in this section describes a Bayesian analysis of survival times of popes, offering a pedagogical example that might effectively appeal to students learning Bayesian modeling for the first time.
Two short technical notes are also part of this issue. The first provides additional discussion on nonecological applications where Taylor’s Law relating the variance and mean of a distribution applies. The second uncovers an equivalence between t-tests and information criteria when doing model selection in regression contexts.
There is one article in the Data Science section that identifies and discusses three important skills to be developed in a graduate-level data science course and, further, discusses desirable characteristics of individuals who would be charged with developing that class.
The November issue concludes with a letter to the editor that complements an earlier TAS article by presenting a particularly lucid derivation for a condition that characterizes when the sign of a regression coefficient will change with the addition of an additional covariate in the model.
Leave a Reply