The training of biostatistical scientists

This concluding session has the ambitious title of Biostatistical Training in the 21st Century. The Biomedical Sciences are changing so rapidly that it would be foolhardy for us to project training needs very much in the future. I am certain that if the biometricians and statisticians contemplating training needs at the turn of the 20th century were alive today they would be stunned as they would not even recognize the proper nouns and verbs that we use every day in our profession. Looking ahead another hundred years from now, the Biostatistical Scientists would look back at our time and comment on the 'quaint' way we planned and analysed studies. They might very well wonder why it was necessary to carry out clinical trials, enlisting hundreds if not thousands of patients, taking many years to complete, when in fact they need only look at a therapeutic drug's chemical composition and predict benet, conditional on knowing an individual's genetic prole. But of course that is a hundred years from now and we must grapple with today's problems. My remarks are targeted at training biostatistical scientists. Not at training biostatisticians. My denition of a biostatistical scientist is an individual trained in statistics, probability, com- puter science and applied mathematics that uses these disciplines to advance our understanding and enlarge our knowledge of important problems in the health and biomedical sciences. Our profession continues to be in a golden period. Mainly because the number of biosta- tistical scientists continues to grow and we are major contributors to the advancement of the health sciences. The major issue in the education of biostatistical scientists is how can we keep the training relevant and make the education meet not only today's needs—but tomorrow's as well. My remarks will be mainly directed at doctoral training—although we would all acknowl- edge that individuals without formal doctoral training could make important contributions. One role model was Jerry Corneld who did not have a doctorate. Sam Greenhouse had made many important contributions to the biomedical sciences long before he received a doctorate. I believe a biostatistical scientist should be broadly trained in statistics, probability, com- puting and applied mathematics. It is possible to be a successful investigator in theoretical statistics by focusing on a very narrow topic. Even a person who is not particularly gifted can make contributions by virtue of long-term contact with the same class of problems. However, a biostatistical scientist must be familiar with a broad range of methodology must have a sound knowledge of theory and must be a capable data analyst. The biostatistical scientist