No announcement yet.

Attention App Developers

  • Filter
  • Time
  • Show
Clear All
new posts

  • Attention App Developers

    I'd long been skeptical of the worth of ethnicity calculators, until I came upon one that impressed me as a doing a (pretty) good job for myself.

    Of course, it's still possible that the relatively good correspondence between my results and the paper trail is just chance.

    But, if there is something more to it, my guess is that it's due to the fortuitous circumstance that the 'ethnic composition' of my actual biological inheritance matches the paper-trail average. That is, while the ordinary variation in DNA contribution from each grandparent in the population at large is significant, just through dumb luck, I may come very close to an even 25%.

    That happenstance could be extremely relevant for the operation of the MDLP Project 4-factor algorithms at Gedmatch. If there were a significant imbalance in the contribution for just two grandparents--say one at 35% and another at 15%, which must routinely occur in reality--the algorithm would be forced to artificially and incorrectly 'average' the ethnic components score in a way that would most likely return nonsense results.

    Given the seemingly trivial differences overall between the component scores among different ethnic groups, the ability of ethnicity calculators to accurately reflect the true level of grandparental contribution must represent the best opportunity for application developers to increase the accuracy of their product's end results. Any developer who crafts their product's interface to allow the user to try a variety of grandparental contribution scenarios in order to achieve a 'best fit' with their known ancestry will have made a major contribution to this field.

    No wonder products like MyOrigins routinely return ridiculous results. As far as I know, they don't even make an effort to place the component scores into the proper context of genealogical transmission.