Estimating the number of groups and group membership using simulation cluster analysis

摘要

A procedure to incorporate measurement error into a cluster analysis through the use of computer simulations is developed. The simulation cluster analysis (SCA) procedure is a method of summarizing the results of the computer simulations in order to estimate the number of groups and the group membership. The SCA procedure requires specification of the measurement error distribution, but initial Monte Carlo studies indicate that precise specification is not necessary. Also, Monte Carlo studies indicate that the SCA procedure performs well in comparison to alternative methods. Two example data sets are analyzed.