Because our groups were based on categorical data, we estimated a series of multinomial logistic regressions to test whether there were differences in regulation or emotionality across the four groups. Odds ratios and confidence intervals are reported. The odds ratio represents the change in the odds that a child is a case (i.e., in the group coded 1 rather than in the group coded 0) corresponding to a 1-unit change in the predictor (Cohen, Cohen, West, & Aiken, 2003) and is often used as a measure of effect size. The number of contrasts was limited to those with the most theoretical relevance to reduce the probability of chance findings and the presentation of results—in our view, those contrasting problem groups with the control group or comparing pure INTs to pure EXTs (i.e., CONT vs. INT, EXT, or CO; INT vs. EXT). First, we set the CONT to be the reference category (this provided three of the four contrasts). Next, we set the INTs to be the reference category to compare INT with EXT (three contrasts were always computed at once,
