The large amount of missing data for fathers is another limitation. We note, however, that there was no significant difference in key variables between adolescents who had missing data for fathers and those without missing data for fathers. Our follow-up analysis indicated that the patterns of associations are largely similar between analysis with the subsample of adolescents having complete parental data and analysis with the whole sample. Thus, our findings were unlikely to be biased due to missing data. Despite the degree of missingness, which we accounted for using multiple imputation, the availability of data for both fathers and mothers allowed us to examine spillover and crossover effects within the family in the context of parental ADS, which also contributes to the relatively small literature on the role of fathers in adolescent development.
