Graduate Education in Computational Statistics

The proliferation of powerful computers and personal work stations has greatly enhanced the ability of both theoryand-methods researchers and data analysts in their practice of statistics. It has increased the availability of statistical software for performing otherwise tedious and time-consuming calculations, and, perhaps more important, computers have opened the way for new analytical techniques. These techniques have been devised by the contemplation of methods that had been impractical previously. Efron (1979) gave a number of specific examples. Some of these techniques have now become standard components of statistical analysis packages. These methods could not have been considered and would not have been developed without the presumption that many data analysts would ultimately have substantial computing power available to them. Not only has computation become an integral part of both the theory and the practice of statistics, but in some academic institutions it is having an equally profound effect on the teaching of statistics [see Eddy et al. (1986) for a discussion of the impact of computers on statistical research]. Consequently, computation should be an integral part of the curriculum at all levels of study in statistics. In many fields, including computer science, active discussion of computing curricula is under way (see, e.g., Koffman, Miller, and Wardle 1984; Koffman, Stemple, and Wardle 1985). Our purpose here is to contribute to the effort of designing appropriate computing courses that may be incorporated in programs of study in statistics. To do so, we have chosen to describe the products of our thinking,