I rather suspect Niklas has a pretty good working knowledge of statistics, given that he worked as an operations research analyst and now teaches mathematics. AFAIK from conversations with him, he never criticized the statistical sampling as too small, and he did not do so in his article (yes, indeed he wrote it). His issue with Overmans starts with his use of the “Totenkartei” and the “Allgemine Kartei”, not the size of the samples, but that the two sources should not be overlapped. It was unclear if Overmans had taken care to not do so. The next issue was Overmans assumption that the "deaths" he believes he found in the “Allgemine Kartei” were in fact dead and that they died as a result of the war. His criticism of the statistical method used is actually:Mori wrote:The Zetterling article has the opposite problem: better knowledge/understanding of OKH documents and other sources but so weak in commenting the statistical approach it ruins its reliability.
"First we have the fact that he has used two different sets of cards to create his sample. If these have identical distributions this causes no difficulty, but there are strong reasons to suspect they are not identical in their distribution over time. Also he has arrived at the overall casualty figure of 5,318,000 dead by adding together the 3.1 million in the “Totenkartei” with the result of a sample from the “Allgemeine Kartei”. This means that the
“Allgemeine Kartei contains, according to Overmans sample, 2.2 million “deaths”. However, since this is a sample it must be added that there is a margin of error on that figure of almost +/- 200,000.22 Since Overmans uses this total as a basis to calculate the losses for smaller periods, we have a case of two statistical errors of margin piled upon each other. If we look at the figures he gives for losses on the eastern front up to 31 August 1943, this means that the margin of error is in the order of +/- 250,000. Consequently, his losses for each month, often given down to the last digit, must be taken with a large grain of salt."
However, that is only one factor in his argument. The second, is:
"...there are many problems with Overmans set of data. First of all, there are the many ambiguities with those cases where there is no information available on the fate of individuals. Overmans equates these cases with deaths. This is far from safe. Overmans argues that the many other possibilities are small. Probably true, but many small factors can taken together be kind significant, if they mainly pull in the same direction. Since
Overmans have assumed that, if there is no information available on the fate of a man after a certain date, he is dead, uncertainties can only pull in one direction. After all, a man can not be more than dead more than once. Hence, there is a basic risk in the kind of investigation performed by Overmans."
BTW, hospital mortality rate for wounded soldiers in the U.S. Army in World War II was 4.5% and they had much better standards of care than the Wehrmacht. With Feldheer WIA of 4,462,965 for 22 June 1941 to 31 March 1945 (OKW KTB, Band IV.2, p. 1515-1516) that is a minimum of 200,833 DOW. Given that the KIA in the same report was 1,149,007, that is a fairly big change.