Phenotypic analyses testing 1) racial/ethnic differences in prevalence and 2) associations between maltreatment and the cannabis measures within race/ethnicity were conducted using SAS v9 (SAS Institute, 2008). Logistic regression analyses examining the association between childhood maltreatment and cannabis initiation and problem use were conducted separately for EA and AA participants using STATA v11 (StataCorp, 2009), with the “robust” option specified to adjust the standard errors for familial clustering. Trivariate twin models examined the relationship between childhood maltreatment, timing of initiation and cannabis problems. Models were fitted to raw ordinal data in Mx (Neale et al., 2003) using full information maximum likelihood estimation. As detailed elsewhere (Neale and Cardon, 1992), twin models allow for the estimation of additive (A) and non-additive (D) genetic factors, shared environmental factors (C; environmental influences that make twins similar to each other), and individual-specific environmental factors (E; environmental influences not shared by twins and error variance). A is indicated when the MZ twin correlation (rMZ) > DZ twin correlation (rDZ), D when rDZ < ½ rMZ, and C when rDZ > ½ rMZ. E is based
