Petraeus fishy numbers

via talking points, a NYTimes article is looking back at the numbers:

Stephen Biddle, a scholar at the nonpartisan Council on Foreign Relations, said Petraeus's December number was "very high" but was likely the result of "statistical noise" — the tendency of Iraq numbers to jump all over the place. Biddle was an adviser to Petraeus last spring but believes the general's testimony was "potentially misleading" because it didn't discuss all the reasons why the numbers might have improved.

the times of course is not fully evaluating their sources:

Biddle was challenged by Lawrence Korb, a senior fellow at the Center for American Progress and a former top Defense Department official in the Reagan administration. Korb said there has been no decrease in violence in Iraq. He noted that August's death toll was higher than July's and pointed to several reports, including a Government Accountability Office study that found no decrease in violence through July. Finally, Korb cited little-reported numbers released by the Pentagon a week after Petraeus testified, which Korb said showed an increase in civilian casualties since the surge began.

and this claim is MORE than fishy:

Petraeus came up with his "over 45 percent" decline by comparing December 2006 and this past August. The December number, in particular, stands out as questionable. For almost all of 2006, the U.S. military count of civilian deaths ran lower than Iraq Body Count's numbers. But the Petraeus number for December, the starting point for measuring the impact of the surge, suddenly leaped 12 percent above the group's, before plunging back well below it.