Originally Posted by
Mr.G
I've never actually seen anyone's results who took it twice. As far as the data, maybe some slight variation is possible, with no-calls or some quality issue, but basically it should be identical. I think the question would be if their algorithm works like Vahaduo, where sometimes even with the same source and target, the result will wiggle back and forth a little with each calculation.
Regarding other aspects of your results, I think I have said before, I'd like to see them work on improving the sensitivity of the communities tool, or to keep growing/developing their samples within the established communities. I still can't fathom how my mom has 87.5% German ancestry, gets a German community -- but my sister and I do not? Yet I do get a Cornwall, England community -- curtesy of one 3x great-grandparent. Weird.
If you feel like it, I'd be curious to see how you would adjust your current percentages to improve your estimate?