February 3 - save the date (or maybe January 8)

Every year at this time since oh, 2004, a group of us have gathered at the County Courthouse to commemorate the US soldiers who have died in the Iraq war. We commemorated the 1000th death, we commemorated the 2000th death, and as I left the crowd the last time I commented darkly, see you next year. I kind of half believed it wouldn't happen, but with the number of US killed now totalling 2950, we seem to be fast approaching the 3000 mark. I collected data on KIAs from the Iraq Coalition Casualty Count website:

Here we see number of soldiers killed, day-by-day, in 2006. Average deaths per day: 2.187. Now suppose we run a regression to find the trend increase in the cumulative count, then use the regression equation to forecast the total number of deaths for the next few months. The regression equation is

Cumulative deaths = 2161.63 + 2.10726*TREND

so we project 2.10726 new deaths each day:

The regression equation predicts we break the 3000 barrier on February 3. But if you look at the data closely, you will notice a small increase in the number of deaths per day. Let's account for this by assuming a quadratic trend. We get:

Cumulative deaths = 2202.925 + 1.399359*TREND + 0.002017*TREND^2

Note the much superior fit. In fact, coefficients on both trend terms are statistically significant at extremely low significance levels. There does indeed seem to be an increasing trend. This equation forecasts the 3000th death to occur on January 8. Happy freakin' New Year.