Bruce A. Craig

In this paper, we extend the concept of a randomized quantile residual to multinomial regression models. Customary diagnostics for these models are limited because they involve difficult‐to‐interpret residuals and often focus on the fit of one category versus the rest. Our residuals account for associations between categories by using the squared Mahalanobis distances of the observed log‐odds relative to their fitted sampling distributions. Aside from sampling variation, these residuals are exactly normal when the data come from the fitted model. This motivates our use of the residuals to detect model misspecification and overdispersion, in addition to an overall goodness‐of‐fit Kolmogorov–Smirnov test. We illustrate the use of the residuals and diagnostics in both simulation and real data studies.

Statisticians with a primary collaborative position at their academic institution have a unique role in combining domain science and statistical research and training and mentoring graduate students. As a result, it is important that their responsibilities be clearly understood by both the collaborative statisticians and their departments. In this paper, we discuss various steps that these statisticians can take to be successful and to be appropriately evaluated for their work. These steps are even more important now because a combination of new technologies, quantitative experts in other fields and the growth of data science in general seems to have put the field of Statistics at a crossroads. For the discipline to continue to flourish, we argue that it is critical that the role of the collaborative statistician be maintained and valued in statistics departments.