Four of the six potential covariates were significantly associated with SOC (p < .05) using a mixed linear model. These covariates were AGP site, verbal/nonverbal status, age of ADI-R completion, and best estimate IQ and together they accounted for 23% of the total variance in the model. The covariates AGP site, verbal/nonverbal status, age of ADI-R completion, and gender were significantly associated with BEH (p < .05) and together accounted for 12% of its total variance. The final set of covariates was selected based on both the strength of association with the subphenotypes and the number of missing values for a particular covariate. For example, best estimate IQ was a very important covariate for SOC in the simple mixed linear model. When covariates verbal/nonverbal status, AGP site, and age of ADI-R completion were included in the multiple model, the association between best estimate IQ and SOC was still significant but with IQ accounting for less than 1% of the total variance of SOC. Since the inclusion of best estimate IQ would reduce the sample size by approximately one quarter due
