STATS.org has an excellent critical analysis of the recent Lancent study that found the statistical mean number of excess post-invasion deaths in Iraq to be around 650,000. The methods used in the study are standard--used routinely and with accuracy by organazations ranging from the U.S. Center for Disease Control, Unicef and the United Nations, and countless government organizations and NGOs throughout the world--and they are pretty solid. (One interesting point that the STATS.org piece doesn't mention is the fact that locales in Iraq which were subjected to heavy or large scale military actions--like Faluja--were purposely excluded in the surveying procedures used in gathering field-data for the Lancet study; had information from such locations been fed into the statistical analysis, the numbers would have simply gone through the roof, as I understand, and what this indicates to me at least is that it is possible that even the highest number of deaths indicated within the study's range is less than the number of excess deaths which have in fact occurred.)

"The Science of Counting the Dead" by Rebecca Goldin, Ph.D.