Trait impulsivity, defined by actions taken without forethought and a consistent preference for immediate over delayed rewards, confers vulnerability to all externalizing spectrum disorders. This includes all disorders along the common developmental progression of attention-deficit/hyperactivity disorder (ADHD) in early childhood to conduct disorder (CD) and delinquency in later childhood and adolescence, to substance use disorders (SUDs) and antisocial personality disorder (ASPD) in adulthood. Such externalizing progression derives from complex interactions among individual-level vulnerabilities and environmental risk factors over time. Specifying how such mechanisms interact across development is a burgeoning area of research. Although trait-level mechanisms have long been studied, research linking trait-level to behavioral mechanisms is more limited. Furthermore, most existing research uses standard inferential approaches, which are not well suited for modeling complex relations among causal influences at different levels of analysis. In this dissertation, I describe how both (1) the methods used to make inference on individual difference correlations across levels of analysis, and (2) the statistical models used to infer how data within levels of analysis arise often fail to fully embody the substantive theories that researchers aim to test. I use my prior work on the “Reliability Paradox” (Haines et al., 2020a) to demonstrate (1), and my work on the Iowa Gambling Task (Haines, Vassileva, & Ahn, 2018) to demonstrate (2). I then discuss a third study (Haines et al., 2020b) that shows how joint generative models across levels of analysis (between behavioral and trait mechanisms, behavioral and neural mechanisms, etc.) can be used to better capture individual differences of theoretical interest.