Where Do We Draw the Line? Assigning Cases to Subsamples for MAMBAC, MAXCOV, and MAXEIG Taxometric Analyses

Abstract

There are several important decisions that must be made when implementing taxometric procedures such as mean above minus below a cut (MAMBAC), maximum covariance (MAXCOV), and maximum eigenvalue (MAXEIG). A Monte Carlo study was performed with 10,000 (5,000 categorical, 5,000 dimensional) samples to examine 5 ways to locate the first and last MAMBAC cuts and 24 ways to perform MAXCOV and MAXEIG. For MAMBAC, there was little difference across conditions, with slightly more accurate results obtained when a small, fixed number of cases ( n = 10 or 25) was located beyond the most extreme cuts. For MAXCOV and MAXEIG, the results were more palpable: MAXCOV slightly outperformed MAXEIG, windows achieved significantly better results than intervals, and a larger number of cases per subsample were associated with more accurate results. Alcohol misuse data obtained from a group of male prisoners were used to illustrate relationships observed in the Monte Carlo study.