Why No Scientist Should Use Classical Inference- Except the Silly Ones

 by Daniel Thorburn

 Research Report 1998:10

 Department of Statistics, Stockholm University, S-106 91 Stockholm, Sweden


Classical statistical inference contains many contradictions and absurdities. We give a collection of short and illustrative examples of various shortcomings. We concentrate on the central issues and not on the problems that may stem from misformulation of the model or the inappropriateness of quantitative methods. The use of classical statistical inference for practical purposes is discussed. In particular we consider it in the context of scientific progress and its use as a basis for decisions. It is shown that the usual output from most inference procedures is of little use. Finally we conclude that the difficulties to use classical inference protects the user from serious consequences of silly mistakes.

Close this Window

Last update: 1998-12-07 / MC