The dependence on noise and inter-subject variability of different group analysis methods in dynamic causal modeling

On a group level, however, several approaches exist: (a) random effects (RFX) second level analysis based only on the mean parameter estimates (b) a more Bayesian-minded approach combining multivariate single subject posterior parameter distributions according to Bayes' theorem (c) temporal averaging where fMRI signals are averaged across the population beforehand. Despite many advantages e.g. avoiding the problem of multiple comparisons, Bayesian second level analysis has only been used sporadically in DCM as a product of Gaussian distributions may lead to counterintuitive results if the participating distributions have differing correlation coefficients (see fig. 1).