The graph is made of data from the Iraq Casu­al­i­ties web­site. To con­trol for dif­fer­ences in month lengths, the data are aver­aged to dai­ly rates per month.

The loess regres­sion is a local­ly weight­ed method, and shows rel­a­tive­ly lit­tle change in the rate at which fatal­i­ties are increas­ing over time.

A while back, Dean Esmay and lies.com got into a thing over how to read this casu­al­ty data. I agree with lies.com that the month­ly data are more appro­pri­ate, but I want­ed to include the last month in the graph above, and I need­ed to weight by day length to make that work. As lies.com notes, there are indi­vid­ual days, and short runs of days that had high­er aver­age fatal­i­ty rates than the ini­tial inva­sion peri­od, and this way of graph­ing con­fus­es the issue.

Dean and lies.com argued about whether there was an upward trend over time (l.c: yes, Dean: no). Dean’s fun­ni­est argu­ment was that the appro­pri­ate regres­sion was a four point mod­el using aggre­gat­ed data for each of four “phas­es” of the oper­a­tion. The prob­lem is, there’s no empir­i­cal jus­ti­fi­ca­tion for those divi­sions. Look at the graph and find inflec­tion points. They aren’t there. And there’s an unam­bigu­ous upward trend. The regres­sion is tricky because of auto­cor­re­la­tion (the fol­low­ing month is like­ly to have sim­i­lar casu­al­ty rates to this month, more sim­i­lar than you’d expect of a ran­dom­ly cho­sen month). Since a lin­ear regres­sion is close to sta­tis­ti­cal sig­nif­i­cance, that mat­ters a lot.

As a con­se­quence, we can­not sub­stan­ti­ate the claim that events like the trans­fer of sov­er­eign­ty (late June, 2004) or the elec­tions (late Jan­u­ary, 2005) made any per­sis­tent, mea­sur­able change in the insur­gency (at least in terms of effec­tive­ness at deliv­er­ing fatalities). 

All of which is very tech­ni­cal and of lim­it­ed interest.

What’s inter­est­ing is that these are a lot of (most­ly) young Amer­i­can men and women who died in Iraq.

They died because the Bush admin­is­tra­tion lied about alu­minum tubes. They died because the Bush admin­is­tra­tion lied about ura­ni­um from Africa. They died clean­ing up a mess that result­ed from poor plan­ning by Don Rumsfeld.

It’s telling that a lot of vet­er­ans are com­ing home from Iraq and run­ning for office. They want a dif­fer­ent direc­tion, and are ready to serve and help. 

If my sources are to be believed, we’ll have anoth­er vet to add to that list soon, tak­ing on Jim Ryun. A vet­er­an vs. a run­ner. I like that fight.

Update: Appar­ent­ly,

These Dems are dan­ger­ous, Trea­so­nous and get­ting away with it. You don’t say a pres­i­dent lied when it was proven he did­n’t lie, and when there is no evi­dence of it dur­ing a time of war. They are aid­ing and abet­ting the enemy.

Also, we “should ques­tion their patri­o­tism. Because they’re act­ing unpatriotically.”

No. Ask­ing these sorts of ques­tions are exact­ly what Con­gress­men are sup­posed to do. It’s called over­sight. While the con­cept has lan­guished in recent years, it’s why Con­gress exists.