University of Washington - Department of Statistics & Psychology
An introduction to the use of permutation and randomization tests in the analysis of randomized experiments as a distribution-free alternative to the normal theory t-test and F-ratio. The talk will touch on:
â€¢ Origins with Fisher (1935) and Pitman (1937).
â€¢ Applicability to experiments based either on random samples or on samples of convenience.
â€¢ Flexibility of choice of error metric and of test statistic.
â€¢ The use of Pitman correlation in the one-way design.
â€¢ The challenge of interaction in the two-factor design.
â€¢ Permutation/randomization based confidence bounds for treatment effects.
â€¢ Some extensions to the multivariate response.
The overheads for the seminar can be found here.