ABSTRACT
The use of significance tests for validating models is basically inadequate, since nonsignificant results only pertain to the compatibility of the data with the model. In this paper an alternative form of data analysis is proposed, whose objective is to assess the acceptability of the model given the data. The use of Bayes-fiducial inference is suggested in this connection; the approach is exemplified through the analysis of an experiment planned for investigating a model of successive stages of information processing in binary choice reaction.
