Linear mixed-model analysis

To analyze the changes in coherence or PAC for selected electrode pairs we used a linear mixed-model approach.⁸⁵ To test for the main and interaction effects of group and recording night, event-triggered coherence and MI values for both nights for groups CON and SCZ were entered into random intercept models, where variability between subjects is accounted for by modelling the participant ID as a random effect (i.e. allowing individuals to have their own intercept about the group mean), and night (baseline vs learning) and group (CON vs SCZ) are added as fixed effects. Models were implemented by using the “lme4” library⁸⁶ in the R environment.⁸⁷ Interaction terms were included if they improved the model significantly (as assessed by a model comparison by using a likelihood ratio test using maximum likelihood estimated versions of the models). If interaction terms were added to the final model, least-squares estimation of group means and differences for each night and post hoc multiple comparisons were calculated by using the “lmerTest” library⁸⁸ where p values were calculated from F statistics of type III by using Satterthwaite approximation of degrees of freedom.