Statistical Tests of Group Differences in ALSCAL-Derived Subject Weights: Some Monte Carlo Results

Abstract

Several techniques to test for group differences in weighted multidimensional scaling (MDS) subject weights have recently been proposed. The present study presents monte carlo results to evaluate the op erating characteristics of two of these with ALSCAL- derived subject weights. The first uses the analysis of angular variation (ANAVA) on raw subject weights. The second applies the analysis of variance (ANOVA) to the flattened subject weights generated by ALSCAL. The ANOVA on flattened weights was less affected by the presence of error and by distortions caused by ALSCAL'S normalization routine than was the ANAVA.