The study found significant differences in the effects of the interaction between a stable propensity for antisocial behavior and the nature of social factors that change over time, depending on the statistical model used. Statistically significant interaction effects were nearly always evident under least-squares-based models. These effects are consistent with models that suggest individual propensity for antisocial behavior tends to magnify the changing risk factors of opportunity, delinquent peer association, and strain, as well as the crime-constraining effects of social bonds. In contrast, tobit models that better match the observed delinquency data only found one significant interaction effect, which was for the measure of delinquent friends. Taken together, the differences across the models heighten concerns that interaction effects in least-squares-based analyses of delinquency could be spurious. This originates from a mismatch between the linear model assumptions and the skewed/censored distribution of the dependent variable. Additional research is required before firm conclusions can be drawn; however, the results of this study should encourage scholars to focus on the adequacy of their statistical models when investigating the manner by which criminological variables interact in their effects on skewed delinquency measures. Data for this study came from the four waves of the Rural Substance Abuse and Violence Project, a prospective longitudinal study of substance use, criminal victimization, and criminal offending among a panel of students in Kentucky. Data were collected initially in the spring of 2001. The dependent variable was a 13-item index of delinquent behavior. Variables related to time-stable antisocial propensity and time-varying delinquency predictors were measured. 4 tables, 1 figure, and 55 references