Usually the PC is used in statistics to do quickly and conveniently what we have always been doing. This is a misuse of the PC since it has the potential to change statistical practice fundamentally. Historically, statistics was developed when computation was hard and expensive. To avoid massive computation a lot of sophisticated theory based on asymptotics was developed. Now computation is cheap and easy. The PC should replace sophisticated theory by simple computations, and make statistics more comprehensible. We should strive for understanding. Sophisticated statistical software is pedagogically harmful. We do not want to solve a dozen problems a day by using recipes. We want to solve a few paradigmatic examples in a few weeks. School statistics should solve a small number of fundamental problems, not quickly, but leisurely. We should derive programs for the solutions, which are general enough, so that they solve a whole class of problems. I do not advocate much deep programming, which is quite difficult. On the other hand, statistical software is for professionals. To learn its use requires an effort comparable to the effort of learning a new programming language. Design of a program is an important part of the learning process. A problem is solved if you have an efficient algorithm for its solution, which you understand.
The CAUSE Research Group is supported in part by a member initiative grant from the American Statistical Association’s Section on Statistics and Data Science Education