Reply to comment

In seeking to explain a statistical anomaly, the first place to look is at the sample. Here, we face a serious selection bias problem.

After all, having experienced nearly a decade of constant deployment, who do you think keeps joining and re-enlisting in the military? I don't imagine it's a random sample of Americans; rather, I imagine the sample skews heavily toward the risk-lovers. These would also be people who are more prone to experiment with drugs. And, lo and behold, we find corroborating evidence in the data regarding rampant drug use -- legal and otherwise -- in the military.

Risk lovers are also people who are disproportionately likely to commit suicide. Case closed? I don't know if selection bias explains the whole phenomenon, but it seems like a good start. As a control group, it would be interesting to follow a sample of people who tried to enlist in the military but were excluded, and track their rates of suicide.

In contrast, evidence shows that the soldiers who are most likely to kill themselves are soldiers who have have ZERO or ONE deployment. Soldiers with multiple deployments are much less likely to do the deed. This undermines the "essential part of what soldiers do these days" thesis.