Originally posted by Paul O'Keefe:
I don't know if which group died in a short time period, the first thousand or the second thousand. It does seem like the rate is accelerating.
It's still an upward trend, but I don't know that it's accelerating.
I did a quick regression and it looks to me like the current trend line is for each month to carry about one more fatality than the month before it.
The formula is
US Monthly fatalities = 46.3+1.03xMonth of conflict where Oct 2005 is month 32.
So for example, we'd expect November fatalities to be about 80, December 81 and so on. Assuming the trend line doesn't change it would imply something like 1060 fatalities for 2006, so we'd pass 3000 around next Thanksgiving.
To see if it were accelerating, I think you'd have to test a number of non linear functions and check the slope at the end of the curve.
I did one gut check and it shows that 6 months ago the simple linear regression showed a slope of 1.24, higher than the current 1.03. That implies a decelleration, but it's not enough evidence to say for sure. I that's the case, although even in that scenario it still implies close to 1000 US dead next year.