Archiver > GENEALOGY-DNA > 2003-11 > 1069786786

Subject: Re: [DNA] DNAPrint on Nightline -- error estimate seen on the certificate
Date: Tue, 25 Nov 2003 13:59:46 EST

In a message dated 11/25/03 9:48:31 AM Pacific Standard Time,

> I thought I remembered something about a 5% error rate, and also that John
> felt the error rate was probably 15 to 20%.

We have to be careful to specify what kind of error we're talking about, and
it's not always clear. The "genotyping" error rate appears to be very low,
even less than 5%. That is, if they run your sample twice, you'll get the same
results for your A-G-C-T readings. That seems plausible, given that we've had a
number of sets of results which are consistent for parent/child relationships.

The bigger numbers (15-20% or whatever) are more like "confidence intervals."
The Maximum Likelihood Estimate, the single most probable point on the
triangle chart, has a lot of uncertaintly. This is partly due to the fuzzy
statistics (there are no markers with alleles occuring in 100% of one population and 0%
of all the other populations). Other factors adding to the uncertainty are
the random inheritance of one or the other allele from your parents and prior
generations, and the difficulty in defining the reference populations (most
populations nowadays have some admixture).

Ann Turner - GENEALOGY-DNA List Administrator
Search or Browse the archives, Subscribe or Unsubscribe at

This thread: