Although tests of model fit indicated that estimating variance components separately for the AA and EA subsamples did not provide a better fit to the data, we are focusing our discussion on the distinctions between the two groups for two reasons. First, there is a paucity of research on differences between African Americans and European Americans in the relative genetic and environmental contributions to alcohol-related phenotypes, so exploratory as they are, our analyses provide a rare window into underlying sources of variance on initiation and problem drinking in African Americans. Second, the fact that the twin models conducted with the full sample roughly mirrored the EA models despite substantial differences in estimates from the separate models suggests that true differences may be masked when European Americans so far outnumber African-Americans in a study sample. Our finding that the best fitting single univariate model of problem use with separate estimates for the AA and EA subsamples was one in which the C estimate for European Americans was 18 or more times higher than the C estimate for African Americans supports this assertion.
