Matlab r2015a statistics toolbox

Statistical analyses were performed using SPSS version 21.0 (IBM Corp.) or Matlab R2015a Statistics Toolbox (Mathworks, Inc.). Primary analyses utilized linear mixed models to compare mGluR5 binding (BP_ND) in the hippocampus and composite association cortex (within-participant factor) between AD and CN groups. The best-fitting variance-covariance structure, as determined by Bayesian information criterion, was compound symmetry. Secondary analyses utilized a similar model with exploratory regions listed in Table 2. Post hoc comparisons utilized unpaired t tests. To evaluate the contribution of GM tissue loss to mGluR5 reductions in AD, group differences in regional BP_ND after GM masking or PVC, as well as in regional GM volume, were also assessed using unpaired t tests. Additional exploratory analyses examined the relationships between hippocampal or association cortical BP_ND and episodic memory (average z-scores for LMII and RAVLT) and global function (CDR sum of boxes [CDR-SB]) in the combined sample with Pearson’s correlation. Tests were two-tailed and used p < 0.05 as a threshold for significance. Vertex-wise, whole cortical analyses were performed with general linear models using Freesurfer. Permutation was used to correct for multiple comparisons. The cluster-forming threshold was p < 0.01, and the cluster-wise threshold was p < 0.05.