The results of using principal coordinates prior to cluster analysis are investigated on the samples from multiple binary outcomes. The retrieval ability of the known clustering algorithm is significantly improved by using principal coordinates instead of using the distance directly transformed from four association coefficients for multiple binary variables.