We used SUDAAN version 10.0.1 (Research Triangle Institute, 2008) to calculate descriptive statistics and Mplus version 6.12 (Muthén and Muthén 1998-2010) to conduct CFA and structural equation modeling (SEM). SUDAAN uses a Taylor series linearization to adjust the standard errors of estimates to take into account the complex sampling methodology of NESARC; Mplus uses a sandwich estimator for this purpose. The weighted least squares estimator with mean and variance correction (WLSMV) is implemented in Mplus, which properly analyzes items with ordinal measurement properties (Lubke and Muthén 2004). Mplus uses all available data during estimation, thus analyses excluded respondents with missing data on all items of a scale but retained participants who had complete data or missing data on one or more items. We used simple random sampling to split the data into halves which we designated “test” and “validation” samples. All analyses were first conducted in the test sample and then verified with the validation sample for replication (i.e. a “split-half approach”). We ran CFA models using the WLSMV estimator then replicated them using the robust maximum likelihood (MLR) estimator.
