<p>Measuring delinquency and victimization during adolescence is critical, given their high prevalence and the strategic importance for violence prevention. Knowledge about the prevalence and risk factors of delinquency and victimization in adolescents relies almost exclusively on self-report studies. Traditionally, such research has been conducted in schools, which allows the inclusion of diverse population groups. However, access to schools remains a significant challenge. This study examines whether school-based probability samples and internet-based non-probability samples are comparable for research on adolescent violence. Using data from 2028 Argentine adolescents (1130 school based and 898 internet based) from the International Self-Report Delinquency Study 4 (ISRD-4), we analyzed selection effects by comparing prevalence rates of both samples and by examining the interaction between survey methodology and theoretical predictors. Results revealed selection effects; the online sample showed lower prevalence rates of violent delinquency and victimization and was composed of more females and adolescents with higher self-control and moral beliefs. The interaction effects showed that the gender-delinquency relationship was significantly stronger in the school sample, suggesting sampling methodology influences this association. While low self-control predicted both delinquency and victimization consistently across samples, low morality specifically predicted delinquency. Assuming external validity of the school sample, these findings provide evidence that although internet-based sampling may produce biased prevalence estimates, it can nevertheless yield valid estimates of the theoretical associations between individual propensities and violent behavior when school access is limited.</p>