Two sources of data were used: Behavior Assessment Unit reports and court case files. The assessment unit reports provided background and clinical information, while the court case files provided information on adjudicative outcomes. Most individuals with Behavioral Assessment Unit reports had court case file information available (208/278; 75%). To assess bias related to the availability of these reports, we examined differences between the subsamples on 40 demographic, historical, clinical, and offense characteristics, using Fisher's exact tests, chi- square tests, or t tests and a Bonferroni correction for multiple comparisons. There were no significant differences between groups on any of the variables, suggesting that data from court case files were missing at random. For logistic models with differing group sizes depending on the outcome examined, patterns of missing data in groups of variables were also examined, and no significant patterns emerged (information available on request from the authors).