Comment by hn_throwaway_99

There are common statistical techniques to better get at causality in this situation. E.g. given how relatively unlikely and random "seeing someone getting shot while still a child" is, it should be fairly easy to match this up with other variables to tease out causality, e.g. just looking at someone in the same socioeconomic situation, same parental situation (i.e. married/single), and then comparing gunshot witnesses vs. others.