Figure 8.

a) Generic path diagram representation of the mediation models used to determine whether vascular and neural variability differentially account for effects of age on rsfMRI variability (where each colour of arrow indicates a separate type of model), plus b) results from one specific model (using variability of time series in Alpha band for rsMEG as a mediator for IC1 of the RSFA). Two of the mediation models aimed to test whether either of the mediators (M) — (1) proxy measures of vascular health (i.e., HRV; Model 1, green paths) or (2) neural variability (i.e., variance of time series for standard frequency bands; Model 2, orange paths)‐explained a significant portion of the shared variance between the independent variable (IV, age) and the dependent variable (DV, rsfMRI variability). The third model (Model 3, blue paths) tested whether HRV mediated the effects of age on MEG neural activity at rest. The relationships between IV‐M and IV‐DV were characterized by path a_i (regression model: M = i₁ + a_iIV+e₁) and path ci (DV = i₂ + c_iIV+e₂), respectively. Path bi described the relationship between M‐DV, while accounting for the effects of IV, path c'_i (DV = i₃ + b_iM + c'_iIV+e₃). Path abi indicated the presence of a significant difference between path c_i and path c'_i. The specific example shown in b) revealed that the ageing effects on rsfMRI variability (in most ICs, see Table 3) were significantly mediated (i.e., significant difference between path c₁ and path c'₁, explaining 48% of the variance) by a summary measure of vascular health (Model 1, green paths). MEG variability in the alpha band mediated the ageing effects on RSFA loading values in IC1 (and IC5, see Table 4), explaining 7% of the variance (Model 2, orange paths), whereas the ageing effects on the MEG variability were not mediated by vascular health measures (Model 3, blue paths). Values beneath paths indicate the Z‐value of the test at a significance level P < 0.05. Percent values beneath paths ab_i indicate the effect size calculated from the proportion between path ab_i and path c_i. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]