PDA

View Full Version : Different results for ancestry composition when uploading the same file



Ann Turner
02-26-2017, 05:52 PM
I seem to recall posts by various people who commented on getting different ancestry results when uploading the identical file several times. My query search terms aren't bringing up any examples. Maybe some of you will recall some examples. This is prompted by the recent television show Inside Edition about identical triplets getting different results.

ArmandoR1b
02-26-2017, 06:15 PM
It starts at http://www.anthrogenica.com/showthread.php?5578-DNA-Land-alternative-to-gedmatch&p=113824&viewfull=1#post113824

I did think of that when I started the thread about the Inside Edition episode at http://www.anthrogenica.com/showthread.php?9822-Inside-Edition-Puts-Ancestry-DNA-Kits-To-The-Test-Among-Sets-Of-Triplets but I decided not to speculate about the FTDNA, 23andme, and AncestryDNA calculators doing the same thing.

Do you know about any identical twins having tested with any of the companies?

firemonkey
02-26-2017, 08:02 PM
I uploaded Ancestry,FTDNA and 23andMe raw data several times. For each the results stayed the same.

cvolt
02-26-2017, 08:23 PM
I re-uploaded my Ancestry results to dna land to see if it'd change, it didn't.

geebee
03-01-2017, 05:21 PM
It actually makes sense that the results may be different, but it shouldn't be hugely different. Ancestry says that it runs the ethnicity test 40 times, breaking your SNPs into slightly different groupings each time. This makes sense if you think about it, since there really is no way to know in advance what constitutes a segment from a given ancestor.

Your score represents the average of the 40 runs, but Ancestry also tells you the range of the results. You might find that different sets of 40 runs may not only give a different average, but slightly different ranges as well.

Likely some people will see little or no difference, especially if their high and low scores weren't too far apart to begin with.

EDIT: I took both the v1 and v2 versions of the test, and there's very little difference in my results. I'm not sure whether to attribute the slight differences -- basically just 2% less Great Britain and 2% more Ireland with v2 -- to the different test, or just to the fact that a new set of 40 runs was involved.

Interestingly, with v1 I had no ethnicity that didn't include 0 on at least one run. Even Great Britain, where the average was 27%. (The highest score over the 40 runs was 53%. Only my Europe West, which had an average of 44%, never produced a 0. (The range was 15-71%.)

For v2, the range for Europe West was nearly identical as with v1: 15-72%. The average was again 44%. As I said, Great Britain was 2% lower at 25%. The range was exactly the same: 0-53%, so I would assume there were a couple more runs that were on the lower side of that.

With v2, there was also one other ethnicity that didn't have a 0 for any of the 40 runs, though it was close. The range for Iberian went from less than 1% (but not 0) to 17%. For v1 was that 0-18%. In both cases, the average ended up as 9%.