These results should be interpreted in the context of the following limitations. First, these results are cross-sectional and therefore cannot speak to the impact of genetic and neural risk on externalizing behaviors over time. In addition, CFAs were used to capture variance shared across externalizing behaviors, creating Externalizing Behavior factor scores. The model fit statistics for theses CFAs were close to, but fell outside the range of a ‘good’ fitting model. The approach also results in factor scores that are specific to the sample in which they were created, decreasing the generalizability of this outcome. Second, while the overall COGA sample is relatively large and diverse, necessary stratification by age and ancestry resulted in some of the analyses being performed in relatively small subgroups. Therefore, it is important that these results be replicated in a larger sample where more complex models (e.g. moderation of association between P3 amplitude and externalizing behavior by the EXT PGS) can be tested. PGS are by nature imprecise as they are an aggregation of variants that are associated with specific behaviors or diagnoses and contain
