Full models including AD and RD as predictors for the performance on neuropsychological tests as criterion measures were found to be nonsignificant in canonical analyses. The analysis of the 5 volumetric measures (set-1) as predictors of the 4 neuropsychological task performances (set-2) yielded four dimensions with standardized canonical correlation coefficients (Rc) of 0.57, 0.45, 0.34, and 0.17, respectively. The full model was statistically significant [Wilks’s λ=0.46 criterion, F(20,170.10)=2.25, p<0.003] and explained about 54% of the variance shared between variable sets. Given the squared canonical correlation coefficients (Rc2) effects for each dimension (0.33, 0.21, 0.12, and 0.03), only Dimension one was considered noteworthy in the context of this study (33% of shared variance). The structure coefficients (Rs) showed that all four neuropsychological performance scores contributed to the synthetic criterion variable (set-2) whereas left HV, left parsorbitalis volume (LParsOrbV), right medial orbitofrontal volume (RMeOFV), and right caudal middle frontal volume (RCMiFV) were the main contributors to the predictor synthetic variable (set-1). The respective squared structure coefficients (Rs2) for both criterion and predictor variables supported these conclusions and are highlighted in bold (Table
