A User's Guide to Misanalyzing Planned Experiments

I have more than 30 years of experience in consulting with research workers on the design, analysis, and interpretation of their experiments. Much of this experience has been with agricultural and biological research workers, but I dare say the abuses and misuses I have found are analogous to any field. It has occurred to me over time that the same errors are being repeated. But what is even more than potentially frightening is the recent availability of very powerful computer software packages on microcomputers, giving people more computing power than was available in mainframe computers only a few short years ago. The wide-spread availability of these powerful microcomputers with very general and equally powerful software packages has its analog in many other areas of life; there is enormous potential for good and for bad. If a user, however dedicated or conscientious, decides to "do his thing" with a set of data, when the researcher is in reality out in "left field," who is going to stop the user from illogical, ill--advised, and incorrect analyses? The researcher may have expert statistical advice available but cannot be bothered to wait a day or two for an appointment to check to see if the model under consideration is appropriate for the design. Or even worse, suppose that no professional statistical advisor is even available within our conscientious researcher's organization. I cannot verify it personally, but I have heard the apocryphal story that some research administrators think that statisticians are now pass with the advent of these all-powerful desk-top computers! Although I have had no personal experience, I would judge that some of the "expert system" computer programs in experimental design would fare no better with naive users.