To test the third hypothesis of moderation by adolescent gender on the association between fathering and adolescent outcomes, the statistical interaction between adolescent gender and father parenting was included in the models. Because failure to specify an existing nonlinear term can lead to spuriously significant interaction terms (Cohen, Cohen, West, & Aiken, 2003), the squared term of fathering was also included. While this term indexed the degree to which there was a nonlinear association between fathering and adolescent outcomes (see Deater-Deckard & Dodge, 1997), we included it primarily as a control to increase our confidence in any significant higher-order interactions with mothering or adolescent gender. Mother parenting, marital quality, gender of adolescent, and family SES were also included as control variables in these analyses.
