accuracy of zetterling s statistics
accuracy of zetterling s statistics
How accurate are the numbers of German losses in Kursk given in the book of Zetterling and Frankson? Are there any NARA divisional and corps verlustmeldungen?
Re: accuracy of zetterling s statistics
I haven't got Zetterling's book, but he seems like a fairly accurate source. Most historians I've read seem to approve of it.
There is a discussion of the book on this thread here.
http://forum.axishistory.com/viewtopic.php?f=79&t=42396
There is a discussion of the book on this thread here.
http://forum.axishistory.com/viewtopic.php?f=79&t=42396
Re: accuracy of zetterling s statistics
All german losses are taken from documents, so they are as accurate as those sources were. And yes they use NARA files but most sources are from BA-MA.teg wrote:How accurate are the numbers of German losses in Kursk given in the book of Zetterling and Frankson? Are there any NARA divisional and corps verlustmeldungen?
Re: accuracy of zetterling s statistics
Undigging an old thread...
I wasn't questionning the figures by Zetterling and Frankson in their "Kursk, 1943" book, but as I checked other documents, I bumped on their source and I couldn't help notice that...
I let you search for the differences.
First image is from the document given as a the source (RH 20-9/441, 10.7.43). Second image is their book. So much for "accuracy".
I wasn't questionning the figures by Zetterling and Frankson in their "Kursk, 1943" book, but as I checked other documents, I bumped on their source and I couldn't help notice that...
I let you search for the differences.
First image is from the document given as a the source (RH 20-9/441, 10.7.43). Second image is their book. So much for "accuracy".
Re: accuracy of zetterling s statistics
You should check out Christopher A. Lawrence book Kursk The battle of Prokhorovka 1,651 pages...well worth the effort and expense.
Re: accuracy of zetterling s statistics
The point above was about German 9th army, that is the northern pincer
I had a very bad experience with Lawrence's War by numbers. The author takes pride of his quantitative effort but doesn't seem to know the most elementary statistics. He fails to ever discuss whether the samples he uses are representative, or what biais they include. He also fails to give sources - as far as I've heard, he belongs to an organization which considers historical data is a trade secret - so that nothing can ever be double checked.
At least with Zetterling & Frankson, I could access the primary documents and spot their typos.
Re: accuracy of zetterling s statistics
These are corrected figures, usually listing the highest figure, e.g a good candidate is the 2nd PzA, the information on their casualties was incomplete and included later in the 10 days reports, at the end of the month (Nachmeldung). If a Nachmeldung was delivered, the figures would be adjusted for one of the three 10 day reports each month (usually dated 4 days after the respective 10 day interval).At least with Zetterling & Frankson, I could access the primary documents and spot their typos.
I do not exclude the possibilty of typos though. The figures in the brackets refer to officers, which are included in the adjacent figure.
The system was far from perfect, but the availability of day to day information was rather unique to the Zitadelle operation, you will rarely find such accurate figures. I guess it may also depend on which level they were compiled.
You can, no doubt a collosal task to gather and compile all information from archive material for all units of all the belligerents, in order to recreate the battles - it is afterall, their intellectual property. I have to agree though, it leaves room for potential manipulations.he belongs to an organization which considers historical data is a trade secret - so that nothing can ever be double checked.
Re: accuracy of zetterling s statistics
If a later report was used instead, the footnote should have clarified the fact ("data for [date] are from [report 1] except for this and this which are taken from [report 2]).
Last edited by Mori on 19 Oct 2020, 11:03, edited 1 time in total.
Re: accuracy of zetterling s statistics
More troubling are other losses report for the exact same dates I found in another folder. These other reports give ca. +15% losses vs the one Zetterling picked in the medical files (because, I suppose, more divisions are accounted for than in the medical file). Either Zetterling did not spot these files in spite of his research - which is all the more annoying since they are in a close by roll and really easy to spot (think 2 pages full of figures)- , or he decided to ignore them - but failed to explain why.Stiltzkin wrote: ↑15 Oct 2020, 06:45These are corrected figures, usually listing the highest figure, e.g a good candidate is the 2nd PzA, the information on their casualties was incomplete and included later in the 10 days reports, at the end of the month (Nachmeldung). If a Nachmeldung was delivered, the figures would be adjusted for one of the three 10 day reports each month (usually dated 4 days after the respective 10 day interval).
I do not exclude the possibilty of typos though.
Taking a higher view, I don't think reconstructing manpower data is possible with numbers down to the individual soldier. In other words, figures with 4 or 5 significant numbers, like we so often read, are nonsense. I can't explain why authors who pride themselves in doing a "statistical" analysis don't use confidence intervals, e.g., "20 000 +/- 10%".
Last edited by Mori on 19 Oct 2020, 11:03, edited 1 time in total.
Re: accuracy of zetterling s statistics
I agree, but read it differently. It's certainly a collosal task to "gather and compile all information": that's the barrier to entry to competitors aiming at the same business niche. Naming the exact source documents wouldn't simplify the work of said competitors.Stiltzkin wrote: ↑15 Oct 2020, 06:45You can, no doubt a collosal task to gather and compile all information from archive material for all units of all the belligerents, in order to recreate the battles - it is afterall, their intellectual property. I have to agree though, it leaves room for potential manipulations.he belongs to an organization which considers historical data is a trade secret - so that nothing can ever be double checked.
It's like saying that a PhD shouldn't mention its sources because anyone would be able to write the same 300+ pages in no time should the NARA box numbers were disclosed.
Re: accuracy of zetterling s statistics
I'm not sure what is the exact case this time but reports can be used for casualty calculations in a combined way (not necessarily one report instead another). Theoretically the numbers from 10.7 could have been adjusted by later reports in a way that certain amount of men were added or taken away for the date 10.7 (without mentioning the the corrected full casualty figure for 10.7). If this is the case here, then I don't see any problem (all the sources for calculations are indicated), if not true, then there is of course a problem.Mori wrote: ↑19 Oct 2020, 10:44If a later report was used instead, the footnote should have clarified the fact ("data for [date] are from [report 1] except for this and this which are taken from [report 2]).
Re: accuracy of zetterling s statistics
Yes, you're correct. I should dig further to check whether further reports correct data from the initial report (although I suspect said reports should first and before all give the 12-20 July and 21-31 July data from Zetterling's table) but I just found this discrepancies while looking for something else. I'm not into Zetterling-bashing of something, I only shared a point I bumped across by chance.
Besides, what I have seen so far in many cases is army administration don't correct previous reports but "fix" figures in the new ones. For ex, if they declare 500 casualties for a period, while it's actually 510, and if they want to fix it somehow, they don't publish a correction to the initial numbers but just add 10 to whatever is reported in the next period.
Besides, what I have seen so far in many cases is army administration don't correct previous reports but "fix" figures in the new ones. For ex, if they declare 500 casualties for a period, while it's actually 510, and if they want to fix it somehow, they don't publish a correction to the initial numbers but just add 10 to whatever is reported in the next period.
Re: accuracy of zetterling s statistics
Indeed, you will never have a perfect picture, but I do not expect massive deviations and manpower generation can be still assessed based on the casualty reports nonetheless, as long as the error is marginal. Soldiers are wounded and may die, soldiers that were missing might be declared dead etc., the reports give a fairly decent overview of the losses a unit has sustained at given instance however. A snapshot, even in an enduring battle. Kursk gives us a good insight into the engagements on the Eastern Front. Envelopments may be a different matter. The error would obviously increase with cummulated losses and small scale contigencies are very volatile.Taking a higher view, I don't think reconstructing manpower data is possible with numbers down to the individual soldier. In other words, figures with 4 or 5 significant numbers, like we so often read, are nonsense. I can't explain why authors who pride themselves in doing a "statistical" analysis don't use confidence intervals, e.g., "20 000 +/- 10%".
Most literature is targeted towards an audience that does not pay attention to intervals. I would definitely include them though (as it is actually a form of quality control), there are however reservations with intervals as well. Most scholars just pick their obligatory range and quite often do not understand the terms or even the system and do not put them into the proper context.
- I have questions
- Member
- Posts: 163
- Joined: 25 Nov 2018, 22:49
- Location: North America
Re: accuracy of zetterling s statistics
But isn't there a risk that the unit keeping the casualty records would understate their own casualties? In that case it would seem impossible to get even a generally accurate picture.