Analytic Approach

To test for group differences in each of the five emotion recognition outcomes (i.e., number of positive and negative labels; negative bias/overextension of negative labels; valence accuracy; label accuracy), we conducted two sets of analyses. First, we examined differences in emotion recognition between children with and without jailed parents using multivariate analysis of variance (MANOVA). Differences between groups were examined with analysis of variance (ANOVA). Second, we tested whether these group differences were maintained with the inclusion of relevant covariates (i.e., child age, race/ethnicity, receipt of special education services, and caregiver’s education) using multivariate analysis of covariance (MANCOVA). Prior to conducting analyses, assumptions for MANOVA and MANCOVA were tested, including basic assumptions (e.g., independence, outliers, linear relationship between independent and dependent variables) along with tests for univariate and multivariate normality, multicollinearity among the dependent variables, and homogeneity of the variance–covariance matrices (Field, 2013; Pepe and Addimando, 2014). Homogeneity of regression slopes is an assumption for MANCOVA, but was not relevant given there were no significant results with the MANCOVA (Field, 2013).