William Kruskal Remembered

Abstract

I knew Bill Kruskal as a dear friend and colleague for over 30 years, but I also knew him as a citizen of his department and university, a statesman of the sta tistics profession and a researcher in mathematical sta tistics. In all of those roles Bill showed characteristics he must have developed at an early age: unshakable in tegrity, consideration for others, painstaking attention to detail and an open, questioning scientific mind. In I hope would be a spirit of social science inquiry that Bill would have sanctioned, I want to begin by ask ing a question of Bill that he asked so often of others. For 30 years, whenever our department met in pri vate session to face a decision on a tenure case, Bill would ask of his colleagues some version of this ques tion: Tell me, he would ask, what specific signifi cant new idea would you associate with the candidate; which of the candidate's works or publications are truly important? Bill's purpose was clear?he wanted fo cus; he did not want to hear a recital of general impres sions, he wanted evidence that would convince him, would convince the dean, would convince the provost and president. I will ask Bill's question about Bill him self, and advance some answers. My first answer is that Bill will be remembered longest for a particular piece of research work during the 1950s. Bill was first appointed as an instructor in our newly formed department in 1950. The best known of his works is the Kruskal-Wallis test, a rank test for the analysis of variance he proposed in 1951 and then developed with Allen Wallis into a famous article pub lished in 1952 (Kruskal and Wallis, 1952). This sim ple procedure has had a remarkable run. If you wish to know the extent of its fame, I suggest visiting Google News, as I did a few times shortly after Bill died. There the name Kruskal produced from 5 to 15 hits on the Google News pages (i.e., the search restricted to news sources of the past month) and almost all of those were