No announcement yet.

Matrix function bug?

  • Filter
  • Time
  • Show
Clear All
new posts

  • Matrix function bug?

    I'm looking right now at the raw segment data downloaded from chromosome browser view. My uncle D has four matching segments on chromosome 2 that all begin at position 8674, the shortest segment being 11.55cM and the longest being 52.34cM (belonging to my mother, D's sister).

    I put these four people into the matrix function to see who matches each other. Amazingly, not a single pair of them matches. The entire table is full of white squares.

    This seems to me to be impossible. Say the first match, match A, matches my uncle on his paternal side. The second match, B, doesn't match A but matches my uncle on the same segment of chromosome 2, so must be a match on his maternal side. But then match C and my mother would have to match either A or B, since C and my mother match my uncle on that same part of chromosome 2 so must match him on one side or the other.

    So it seems people are being incorrectly reported as non-matches. I can think of several reasons this may be happening, but I'd like to know definitively since then possibly I could come up with a workaround.

    1) Does the matrix function use the same threshold to determine if two people match as Family Finder generally? The overlap of these segments is over 11cM, so every pair of these people who actually does match should be matching strongly enough to be classified as a match by Family Finder. But if the matrix function sets a higher threshold, that could explain what I'm seeing.
    2) Could this be related to the bug with the search feature reported here?
    The search function effectively sometimes shows some weak matches as non-matches by not showing them in a search of the surname
    3) My kits show a high percentage of no-calls. There is reason to believe that all recent FTDNA testers are showing a high percentage because FTDNA stopped measuring 28000 or so SNPs but still reports them as no-calls in the raw data file -- see this post But if my uncle's kit really does have too many no-calls, he might not actually be a match to these people, and that's showing up in the fact that these people don't match each other.
    Last edited by barryrsmith; 4 September 2018, 08:03 AM.