Daniel R. Jeske
The August 2019 issue of The American Statistician features a dozen articles spanning a variety of interesting topics.
The General section has five articles. The first shows how a simple transformation of a test statistic can increase its power when it is used in a hypothesis testing context. Illustrations of the idea include a change point model setting. The second article discusses disease mapping. A distinction between generative models and internally standardized models is made, and the ideas are illustrated by analyzing a Ohio lung cancer data set. The third article brings additional focus to the p-value topic. Comparisons and contrasts of frequentist and Bayesian foundations are offered. The fourth examines factors that go into why statistics is selected as an occupational choice. Certain personality characteristics are among the important factors identified. The final article in the section discusses the role of Wikipedia in learning about statistics. A tool is provided for evaluating the quality of Wikipedia articles.
The Statistical Practice section has one article this month, and it discusses how the ways in which a survey is designed or presented to respondents can have an effect on the results obtained. A new approach for analyzing surveys characterized by these so-called framing effects is presented.
There are four articles in the Teacher’s Corner. In the first article, a novel approach for sharpening Jensen’s inequality is presented with comparisons to related ideas. The second article shows how to simulate a bivariate data set with arbitrarily specified marginals and a specified (subject to constraints) correlation. In the third article, two continuous mixture models, beta-geometric and gamma-exponential, are compared and shown to be similar when used to model data. The final article in this section outlines a sequence of guided experiments with a gumball machine to teach about the concepts of hypothesis testing.
The issue has single articles in the Data Science and Short Technical Note sections. The data science paper presents a new method for jointly clustering rows and columns of a matrix. Comparisons are made to existing methods. The short technical note reveals a problem with the R-squared formula when Bayesian regression is employed, and then proposes a new definition of R-squared to fix the problem.
The August issue closes with a brief Letter to the Editor providing an additional example of aliasing that arises when modeling disease severity.
To read this issue or learn about how to submit an article, visit The American Statistician, website.
Leave a Reply