PDA

View Full Version : Why are D-stats of same samples differing between datasets?



Bas
11-05-2017, 06:55 PM
Running this D-stat with a fresh Reich lab EIG file and my own merged one, I get:



Reich lab (untouched): Chimp Hungary_HG Hungary_N Iberia_HG Z= 18 +/-
2 Reich lab sets merged in PLINK: Chimp Hungary_HG Hungary_N Iberia_HG Z= 1.1 +/-

Does anyone have any ideas on why exactly the same samples, with exactly the same number of individuals and roughly the same number of SNPS in both runs should differ so much?

Chad Rohlfsen
11-05-2017, 06:57 PM
Need more info.. what sets and any errors during merge?

Bas
11-05-2017, 07:00 PM
Need more info.. what sets and any errors during merge?

The first one, with a Z-score of 18 (which I probably trust more) is the Minoan and Mycenean dataset. The second one is a merge of that with Haak.

No errors during merge, but did have to remove common samples as there is quite a bit of overlap. Didn't do anything to the SNPS though.

Chad Rohlfsen
11-05-2017, 07:05 PM
Why would you merge Haak with that?

Bas
11-05-2017, 07:07 PM
Why would you merge Haak with that?

The Minoan and Mycenean had Rathlin, which admittedly was the main attraction but the Haak dataset also has moderns whereas the Minoan is only Ancients.

Chad Rohlfsen
11-05-2017, 07:20 PM
But, why Haak? I think that is only in Affyx, so it won't merge with the other one without other steps. Why not just get the Near East Public set and add the Mycenaean and Minoan samples? Stick with the 1,240K stuff. Haak is pretty useless.. a lot of those samples have been redone with more coverage. Haak is only like 300k anyway.. I'm not sure why you're messing with that.

Bas
11-05-2017, 07:32 PM
But, why Haak? I think that is only in Affyx, so it won't merge with the other one without other steps. Why not just get the Near East Public set and add the Mycenaean and Minoan samples? Stick with the 1,240K stuff. Haak is pretty useless.. a lot of those samples have been redone with more coverage. Haak is only like 300k anyway.. I'm not sure why you're messing with that.

That is probably the problem! I hadn't looked inside the Near East set to see exactly what samples were in there. Cheers Chad. Bit of a newbie to ADMIXTOOLS. Going to merge them now.

Chad Rohlfsen
11-05-2017, 07:41 PM
Alright. Any issues, just reply here or send a message.