Sens. Sam Brownback and Pat Roberts have written a letter to the Department of Veterans Affairs urging full funding for a new cemetery at Fort Riley in Kansas.

The reason: With an influx of casualties from Iraq, the existing cemetery at Fort Riley is now full. Well, not entirely full: A spokesman for the facility tells Reuters that bodies can be buried on top of other bodies if family members want to share plots.

Thank heavens we have such stalwart men looking out for our troops. It would be horrific to force families to have to share their grieving space with other people.

Rather than building more cemeteries, there is another option available to a Senator ? working to make sure that fewer soldier die in Iraq. Alas, Roberts and Brownback have steadfastly opposed any Congressional move that would hasten our exit from Iraq, just as they opposed any moves that might have kept us out in the first place. Luckily, Brownback seems likely to stick to his promise not to run for re-election, and it looks like former Congressman Jim Slattery will be running against Pat Roberts.

While we’re on the topic, here’s the mighty surge at work. (Note that, because I fixed a small bug in the code which generates this graph, the linear regression line is higher than it has been in past graphs. Older graphs plotted it too low because I was converting the average numbers to an integer, which threw away the decimal portion of the average, systematically lowering the number. This graph is correct, and the others underestimated the intercept of the linear trend, though the slope would have been basically unaffected. I never could understand why the linear regression was always so much lower than the loess regression.)

The pink area of the graph represents a bootstrapped loess regression, 500 of the red lines generated with only half (or 95%) of the data points. It represents an estimate of the underlying variability of the data and the plausible range of the actual trend. To my eye, it shows an upward curve since February, an accelerating rate of fatalities, rather than a flattening.

Since the escalation in Iraq (the “surge”), half the months have seen more fatalities than the linear regression would predict, half have been below the trendlines, exactly as would be expected if no change had happened. While I certainly hope that there is a signal in those last three points, there is no statistical basis for that conclusion, especially given the apparently accelerating trend in the locally-weighted regressions.