"Here is a something from Dienekes using Iraqi Kurdish samples from Xings et al.
I heard this argument a lot. Let me briefly explain why this post of Dienekes is a bad example/evidence.
There is a very good reason why Dienekes never used the Xing et al dataset for comparison again. Kurti was referring to Dienekes' plot which is very noisy caused by the low number of compared SNPs.
It messed up the position of Kurds in the plot and all other populations in the plot. For example, just take a look at the Turks merging with Armenians in the plot. We now know this is not correct. There are major differences but the plot with the low K value could not see it in the relatively low number of SNPs. Dienekes never repeated any analysis with Xing et al because of that. However, Razib Khan, another science blogger started to use this noisy plot of Dienekes for his simplistic argument that Turks are just assimilated Armenians.
The Xing et al data are based on 250,000SNPs, but most of these SNPs are not part of other genetic studies, so a comparison is nearly impossible. Zack from the Harappa Ancestry project is trying to use Xing et al., but he has to reduce the number of SNPs down to 29,000 SNPs!
"This dataset is valuable because it contains several South Asian, Central Asian, Southeast Asian and Caucasian groups. However, it does not have a good SNP overlap with 23andme and the other datasets. It has only about 29,000 SNPs in common with 23andme v2 data. Combining HapMap, HGDP, SGVP, Behar et al and Xing et al with 23andme data leaves us with 25,000 SNPs. Due to that, I'll be using Xing et al data for only a few analyses."Zack also wrote:
"Do note that the Xing results were computed with a smaller number of SNPs and thus might be noisy."