It's a cute little graph, but it is meaningless. What you have with Bush is an accelerating trend in hostile casualties. Yes, I understand that you would like me to see the data the way you framed it, but the way you are framing it makes no sense.
If you look at the Clinton years, there are considerably more people in the military than now, there are very few hostile deaths, and the proportions of other deaths are about the same. It is always the proportion that is important, and the rate of change in a particular category. If you look, for example, at the rate of accident, when there were more men in the military you have more deaths by accident, and as the numbers of men in service fall the number of deaths decreases, not because it got "safer" under Clinton- though using your kind of graph I could certainly make it look that way- but because there were fewer men, and thus a smaller number of deaths by accident but roughly the same percentage of deaths by accident. Usuing your kind of graph I could totally hide the fact that I was pulling my numbers from different base sizes of men. It's a clever way to fake statistical data, and a reason you need to look at all the columns.
So what does this mean for you?
For your graph to be meaningful you would need to take the total men and find the percentage who died out of that total, and show what the comparable percentages are. I'm just eyeballing it, but as the troop numbers go up, as you go back in to the 90's, the deaths go up, as you would expect, since you have more men.
For anyone looking at statistics the most important thing is rate of change. I understand why you want to ignore the rate of change issue, since it plays so badly for you. But telling me I have a reading comprehension problem when I tell you I see a problematic rate of change problem, is not going to convince anyone but the most math challenged individuals that you have a point. |
