To account for non-normality in our data (a typical phenomenon in sex research), we used the Robust Maximum Likelihood (MLR) estimator, which corrects for deviation from multivariate normality by
computing robust standard errors and an adjusted chi - square (Sass, Schmitt, & Marsh, 2014).
Where trials have involved the randomisation of clusters, we anticipate that study investigators will have presented their results after appropriately controlling for clustering effects (
using robust standard errors or hierarchical linear models).
Robust standard errors are reported in parentheses.
This was tested using logistic regression (with
robust standard errors) to model the extent to which wave 1 demographic characteristics and outcomes affect the probability of a family participating at wave 3.
We apply Linear Probability Models (with
robust standard errors) for the analyses on both outcomes.6
Entering both moderators in one model with
robust standard errors did not change the study findings; entering both moderators in one model without robust standard errors slightly changed the study findings, such that only children's emotional problems remained a significant moderator.