Randomization Test of Paired Data: Application to Evoked Responses
Donald C. Martin, Veronica Buffington, Joseph Becker · Psychophysiology · 1981
ABSTRACT This paper describes a randomization test for similarities in observation vectors between pairs of subjects. Visual evoked response data for pairs of relatives were used in an example of the test. The randomization test does not require any distributional assumptions and can be applied to both univariate and multivariate cases. The basic procedure has three steps. First, some measure of similarity between subject pairs is defined. Second, the distribution of average similarity under random pairings is found. Third, if the average similarity for observed pairs is in the tail of the distribution generated by the assumption of random pairings, the hypothesis of no greater than chance similarity is rejected. The distribution of average similarity under the assumption of random pairings is approximated using a permutation central limit theorem. The randomization test can be used instead of Barttlet's test for nonzero canonical correlations.