Discussion of "The Need for More Emphasis on Prediction: A 'Nondenominational' Model-Based Approach"

David Harville has provided a compelling case for an increased focus on prediction in the teaching of statistics. I am very sympathetic to Harville’s plea. Indeed, it was the ability of statistical methods to address prediction problems (in sports and finance, the same fields that Harville mentions) that attracted me to the field of statistics more than 30 years ago. Even in application areas where the focus has been on parameter estimation, e.g., regression coefficients in economics or treatment effects in clinical trials in medicine, it seems quite natural to me to think of these parameter estimates in terms of the predictions that they imply. Given that I agree with Harville on the central role of prediction, my comments below concern his focus on models and the relevance of the “nondenominational” approach.