As the moderator effects pertain to pre-registered a-priori hypotheses, I would go ahead and test and report these effects (preferably using likelihood ratio tests based on model comparisons), even though the variance of slopes may not be significant. 6.
← all excerpts
The neural determinants of age-related changes in fluid intelligence: a pre-registered, longitudinal analysis in UK Biobank.
3
—
—
The sentences
Moreover, in the case of highly collinear predictors, this can artificially increase the difference between the predictors (with one tract highly significant, the other non-significant) merely because they have very similar predictions.
To do so, we refit the above MIMIC model, including only those white and grey matter regions that were nominally significant in the modality-specific analyses.