MODEL VALIDATION BY PERMUTATION TESTS - APPLICATIONS TO VARIABLE SELECTION

Regression model validation by permutation tests was explored. Especia lly in cases where the model significance is doubtful, a permutation t est adds crucial information which can often can be decisive for the e xistence of the model. The background and applicability of the test pr ocedure are described. As an example, the use of permutation tests was extended to validation and investigation of four predictor variable s election techniques, namely MUSEUM, GOLPE, VEP and IVS-PLS. The select ion methods are briefly reviewed and compared. The permutation tests w ere applied before, during and after variable selection. Some similari ties and differences in the behaviour of the variable selection techni ques were found and are commented upon. (C) 1996 by John Wiley & Sons, Ltd.