Tollense Valley Bronze Age battle field, standardized PCA-results

Using the same standardized data we have the following PCA plot, which differs from what we see on plots made using only partly overlapping SNP sets.  I don't see any reason to use Mediterranean samples, because of the small SNP number of some samples.  What we see in general on the plot is that most ancient samples  fall between Germans and Poles.  We see also that Finns, Russians, Poles and Norwegians show genetic drift.  The most Polish ancient sample is WEZ56 and WEZ54 falls inside the British cluster.  Samples WEZ39, WEZ40 and WEZ51 fall somewhat closer Finns, being still Central European. WEZ56 is the most Polish sample in the original study graphics too.

Tollense Valley Bronze Age battle field, standardized F3-results

According to my experience the f3-analysis (and dstat) generates error due to differences in SNP numbers between individuals.  Because the SNP number can vary as to the sample source, I removed all "bad" SNPs from the study data and added Finnish samples to make it equal to other sample groups.   Actually the average SNP amount in individual tests didn't change much from the original situation, with exception of samples gathered from the 1000genomes project.  After this operation the SNP number in each test between ancient and modern samples was almost constant.

Average SNP numbers per ancient sample, the difference inside one test group a few hundreds in maximum

WEZ15    56606
WEZ16    5453
WEZ24    13749
WEZ35-2    28766
WEZ39    9343
WEZ40    21711
WEZ48    6392
WEZ51    15313
WEZ53    14758
WEZ54    29468
WEZ56    28034
WEZ57    34161
WEZ58    21152
WEZ59    28256
WEZ61    34657
WEZ63    11920
WEZ64-1    26150
WEZ71    15698
WEZ74    9891
WEZ77    15721
WEZ83    14999

 Result of f3-tests using Mbuti as an outgroup:

All data with exception of Finnish samples (1000genomes) are from the study

Addressing Challenges of Ancient DNASequence Data Obtained with NextGeneration Methods