This is a cross-posting from a thread on 23andMe about looking at small segments. ==== OK, here's a case study about a small segment. I was recently contacted by a Family Finder match (LM). She and her half-first cousin once removed believe they descend from an undocumented WRIGHT ancestor. Both shared a segment with my sister (ME) and noticed that we also have a WRIGHT ancestor. Furthermore, LM matched my second cousin once removed (JJ), with the longest segment on a different chromosome. The common ancestor for JJ and ME is a WRIGHT. That looks promising. I looked at the smaller segments, and lo and behold, there's one short segment on chromosome 2 that is found in all three pairwise comparisons. The numbers are chromosome, start position, stop position, cM, and number of SNPs. ME & JJ 2 135,763,071 138,614,305 2.36 700 ME & LM 2 135,135,633 138,614,305 2.44 800 JJ & LM 2 135,763,071 138,614,305 2.36 700 The apparent exactness of the boundaries is an artifact of the way FTDNA analyzes data in blocks of 100 SNPs. LM sent me her raw data, so I could use David Pike's utility to get closer to the actual boundaries. There is still some fuzziness, but you can see that the actual segment is somewhat longer than reported by FTDNA. ME & LM 134,827,563 138,652,870 897 [Rutger's map interpolator gives this distance as 2.83 cM] When I compare LM with 23andMe v2 data for ME, the SNP count drops since there is not complete overlap in SNPs tested, but the boundaries are similar. ME & LM 134,827,909 138,625,822 384 I also have other WRIGHT connections at 23andMe: myself (AT), my son (JT), and my uncle (JW). LM & JT 135,259,837 138,401,572 702 [v3] LM & AT 135,153,875 138,459,636 320 [v2, so fewer SNPs, but boundaries are similar to son JT] LM & JW 134,827,909 138,576,971 376 [v2] That segment seemed to be holding up well with all those pair-wise matches. It's encouraging that the match in my son appears to be inherited from me. BUT -- JT and I do not match my sister (or JW or JJ) in that region! The fact that the same segment showed up in child JT and parent AT was not a sufficient filter to rule out a false positive (what I call a pseudo-segment). The reason for the false positive is clear when we can use phased data. I have father/mother/child data for JT, so I can phase his genotype and use the alleles he inherited from me to compare with LM. What appears to be one long continuous run of half-identical SNPs was actually ten shorter runs stitched together with some alleles from his father, which made it appear the run was continuing. It takes many pairwise matches and/or phased data to be confident that everyone shares the same segment IBD when you're looking at short segments. SageGrouse is looking at 10 cM matches, and I agree that those are long enough to rule out pseudo-segments, provided you have some minimum number of cousins to compare. I looked at some data I have for my family study of an inherited hearing impairment. There is a segment which must be IBD because we share the phenotype. In that case, four cousins performed as well as a father/mother/child trio.