In a longitudinal study of nicotine vaping’s influence on cannabis initiation, we found that the positivity causal identification assumption was violated. Regardless, we still estimated the average treatment effect (ATE) and an analogue estimand that does not require the positivity assumption. The analogue, the incremental propensity score intervention (IPSI), was the difference in cannabis initiation risks had everyone’s odds of nicotine vaping been decreased up to 90% compared with observed nicotine vaping odds. Interpetations of results from the IPSI and the ATE were consistent with each other. That is, lower odds of nicotine vaping decreases cannabis initiation risks (IPSI), and nicotine vaping increases cannabis initiation risk (ATE). Researchers should consider using shift estimands like IPSI when the positivity assumption is violated.
