was slightly improved. In the first study, we accurately classified 88% of FAS and 68% of AE/Non-FAS subjects. In the current study, these classification rates were 77% and 70%, respectively. There were several significant differences between the two studies that could account for these results, such as included sites and the variables chosen for analysis. In the first study (CIFASD I), just two sites were included: San Diego and Helsinki, Finland. In the current study (CIFASD II), data were collected from five U.S. sites and one South African site, which increased the heterogeneity of the sample. The increase in sample diversity related to changing sites likely increased the variability of neuropsychological test performance and may have lead to lower classification rates. In addition, a different strategy was used for selecting the neuropsychological variables for analysis. In the first study, we chose variables based on their ability to differentiate the AE from CON subjects in univariate analyses. In the current study, we used clinical judgment and chose the traditional variables from each measure. Although the effect sizes for these variables were mostly in the large range, the inclusion of additional or different variables may have improved our classification accuracy. In the
