Date
1  3 of 3
False negatives in Family Finder test #dna
Gary Mokotoff <mokotoff@...>
It is my understanding that the Family Finder test of Family Tree DNA can
produce false positives, that is, two people may show close kinship but actually may not be that closely related. My question is whether the Family Finder test can also produce false negatives. Specifically, I tested a woman and a man that hopefully was her grandnephew (her sister's son's son). There was no Family Finder match. Does this unconditionally conclude that they are not related? If false negatives do occur, does anyone know at what rate given the kinship described above? Gary Mokotoff


Judy Simon
There is a lot of randomness in the process of passing autosomal
toggle quoted messageShow quoted text
chromosome segments to the next generation; by the luck of the draw, it is possible for a woman and her sister's grandson to have no segments in common in the Family Finder test, yielding a false negative test. I don't think this this situation is common, but I don't have enough knowledge and experience to put a probability on it. There are others on the SIG who know more than I do, so hopefully someone will come up with an estimate. Judy Simon Stony Brook, NY
On Sat, Jul 21, 2012 at 6:12 PM, Gary Mokotoff <mokotoff@earthlink.net> wrote:


Sarah L Meyer
The answer to the probability question is in part given in the FAQ's (What
toggle quoted messageShow quoted text
is the probability that my relative and I share enough DNA for Family Finder to detect? The FAQ's have chance of detection 2nd cousin or closer >.99 3rd cousins >.90 4th cousins >.5 5th cousin >.1 6th and more distant <.02 I had a second cousin once removed, test as a distant cousin, so my interpretation is that that these percentages are the chances of detecting the correct relationship (and not just any relationship). Unless FTDNA is saying that those probabilities are the probability of detection at any level I have to assume that they are at the stated level. So the probability of being missed is 1 the probability of being detected. So for second cousins or closer there is less than a 1% chance of being missed and for third cousins is <10%, fourth cousins <50 percent etc. (BTW Gedmatch had him at 3.5 generations to MRCA). Ignoring the less than or greater than signs (since they did not give us more places past the decimal), in my case without considering the possibility of a second cousin being considered a 3rd cousin etc with TWO people being missed, the chances of missing both were (less than) .0001 or 1 in 10,000, which is why if there was no error in the DNA processing, I was going to look for "the milkman". And if I were to take into consideration  a match at *any* level, the chances would have been much smaller. In this case, I think we need to look at this as if they were second cousins (they are actually closer) and the chances that they did not find them as second cousins is less than 1%, the chance that they will be found as second cousins or third cousins (but not second) is .99+.01*.9=.999 so the chance of missing them as both a second cousin and a third cousin would be .001 etc. A calculations gives that the probability of detection at any level up through 5th cousins is .99955 which would say the probability assuming that each of those > signs was an equal sign and that the match did not occur at any level is .00045. This does not even take into account the probability of a 6th or greater cousin match. That is why at this level of known relationship, I suggested that GEDMATCH.com be used to see if the location of the error could be determined. Sarah L M Christiansen Georgetown TX 78633
From: Judy Simon <heyjude0701@gmail.com>

