The goal of the current study was to examine the relationship between the positive illusory bias and the subtypes of aggression in a sample of aggressive children (n = 120) across three time points using multiple informants. In addition, a methodological question was explored by using difference and residual scores of the bias to examine its relationship with proactive and reactive aggression. While difference scores have traditionally been calculated to represent the bias, more recently, residual scores have been used (White & Kistner, 2011). To date, no study has examined whether unique findings emerge if the bias score is represented by a difference versus a residual score. To address these goals, four cross-lagged autoregressive models were estimated, including teacher and peer-informed difference and residual models. Finally, the relationship between the positive illusory bias and specific social cognitive biases related to the subtypes of aggression was examined. Results from the teacher-informed difference model indicate that the positive illusory bias predicts proactive and reactive aggression across time controlling for within-time associations among constructs. In this sample, the residual scores did not seem to accurately represent the bias, and therefore, there was no support for using a residual score to examine the positive illusory bias. Lastly, the hostile attribution bias, a social cognitive bias associated with reactive aggression, was found to predict the positive illusory bias during elementary school, but not during transition to middle school.