The reason we haven't parsed it in any more detail is that statistically we don't have sufficient power. We can only really look at one variable at a time, such as deployment, yes or no; or army or non-army, and those sorts of things. If we start looking at what's called a bivariate analysis, looking at two variables at a time, we find ourselves not being able to say anything because there's no statistical power to back it up.
On March 20th, 2017. See this statement in context.