Friday, 10 April 2015

More on "New Ancestor Discoveries" through AncestryDNA

Last Saturday in a blog post I commented on this new AncestryDNA feature pointing out that, in contrast to the usual hype from the Ancestry PR machine Kenny Freestone, AncestryDNA’s Director of Product, had produced a video stating to expect that:

about one-third of subscribers will have no new ancestors identified,
about one-quarter of the matches will be false.
Now he's posted on the Ancestry blog New Ancestor Discoveries: Clues (Not Proof) to Your Past in an attempt to"explain more clearly what this feature is and is not." It's welcome, addresses the questions:
What is a New Ancestor Discovery?
Why do we think you are related to this person?
What is the confidence you are really related to this person?
How is the confidence determined?
and includes the statement, pretty much in line with the statistic quoted previously:
In general, the confidence that a New Ancestor Discovery really fits in your family tree is pretty good—about 70%.
I have no way to independently evaluate whether about 70% is an accurate figure, although if it is it's for relatives, not ancestors.

If you've had experience with Ancestry's "shaky leaves" on family trees you'll appreciate that they are often wildly off. My last two were for people in a different continent, even cursory examination showed them to be false - false positives.
With statistical techniques you have to choose the balance been false positives, giving too many leads that are incorrect, and false negatives, not giving leads which would have been helpful.
With the shaky leaves, as with the "New Ancestor Discoveries" Ancestry has evidently chosen the former, perhaps because it gives the appearance of working harder for the client. I wonder if this has been focus group tested?
Explanations as in the blog post help, but Ancestry could do more. One relatively simple thing would be to stop referring to relatives as ancestors. It only alienates. Beyond that fire the person who refuses to give us a chromosome browser, continue to work on improving the technique, and put an enhanced effort into expanding the database. Not charging so much more for the service in the UK than the USA would help.

No comments: