The data were also analyzed by constructing age- and education-adjusted Z-scores based on our own control group. First, within the control group, regression analysis on raw scores for each cognitive measure (including both the MicroCog and supplementary measures) was carried out to determine coefficients to correct for age and education. These coefficients were then applied to all subjects’ data (both control and TNAD) to compute predicted scores and residuals based on the association of age and education with the measure. Next, for all subjects, the residuals were added to the NAC raw score mean of the measure to obtain scores adjusted for the association of age and education (in controls) with the measure. Finally, Z-scores for both TNAD and NAC were computed using means and standard deviations of these adjusted scores from NAC. Using the adjusted Z-scores based on controls, multivariate analysis of variance (MANOVA) was carried out, with Group and Gender as main effects, and average Z-scores for the 9 domains as dependent variables, and a Group by Gender ANOVA was performed for the average Z-score (based on controls, age- and education-adjusted) across all domains.
