No announcement yet.

Family Finder relationship estimation

  • Filter
  • Time
  • Show
Clear All
new posts

  • Family Finder relationship estimation

    Does anyone else find the Family Finder estimation of relationships to be sometimes very far off? For instance, there's one of my matches that comes up as a 2nd-4th cousin with 45 cMs of shared DNA and a longest block of 28 and has a very complete family tree. When I compare our trees and count the generation, it works out to more like 7th cousins. This is not the only example I've found but some others also have extensive Great Migration (early New England) ancestry where interbreeding might account for overestimation of the mutual relationship.

  • #2
    Absolutely! Two points here: First, the FTDNA relationship estimates of 2nd to 4th cousins and beyond seem to be excessively optimistic most of the time. Very, very few of my matches in that range can be connected to my very deep pedigree, and of the few I can identify, most appear to be at least one or two generations farther back than estimated.

    Second, the estimates at GEDmatch, where we get a single value (say, 4.0 for a 3rd cousin), are usually very close to the mark out to about 3rd cousins, but beyond that, they become increasingly "off". Further, even if the estimate reflects something like the theoretical relationship level for a given level of autosomal matching, the underlying distribution is strongly skewed. This is a consequence of the way recombination and the random assortment of chromosomes works; I don't see how any algorithm could possibly compensate. The solution would seem to be to provide better labels (FTDNA) or better guidance (GEDmatch) for the interpretation of the estimated degree of relationship.


    • #3
      I recently viewed a webinar by "The Genetic Genealogist" blogger, Blaine Bettinger. In it, he gave some advice on how to use the various DNA testing company's estimates of relationship.

      For FTDNA, he advised selecting a match, then view the matching segments for that match using the Chromosome Browser (you can use the link for "View this data in a table"). Next, remove ALL segments smaller than 5cM. Add up what is left, and use one of the charts that show cM estimates for various relationships to then see what the relationship could be (such as the Shared cM Project chart at ISOGG, or the "DNA Detectives" Autosomal Statistics Chart at "The DNA Geek's" blog.

      Individual situations such as endogamy (marrying within a specific social group, caste or ethnic group), or having parents/grandparents related, can affect the relationship estimates as well.


      • #4
        I just found another example. Someone who is supposed to be a "2nd-4th cousin" (51 cM shared) who has a very detailed tree and the only connection I can find is in the 1600s.