GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2008-01 > 1200174432


From: "Ken Nordtvedt" <>
Subject: Re: [DNA] Markers - How many and which ones?
Date: Sat, 12 Jan 2008 14:47:12 -0700
References: <2dff56a0801121109k2ae8145ai6ca2b79ec374e52f@mail.gmail.com><D1A1FA1F-915D-4DFF-B592-C958D8E7E8A6@vizachero.com>


----- Original Message -----
From: "Vincent Vizachero" <>
To: <>
Sent: Saturday, January 12, 2008 2:40 PM
Subject: Re: [DNA] Markers - How many and which ones?


> Those who have problems accepting the utility of TMRCA estimates
> often fail to accept them for what they are, which to say "estimates".
>
> Also, though we often adopt the sloppy habit of quoting point
> estimates (usually the 50% probability point, but sometimes the MLE)
> we should always keep in mind that the number we are using is
> implicitly a confidence interval.
>
> If a 50% success rate is good for you, then just refer to the 50%
> confidence interval.
>
> For example, assume you have two haplotypes with GD of 3 for 37
> tested markers with an average mutation rate of 0.0049.
>
> The most likely estimate of generations seperating the two births is
> 17 generations (i.e. TMRCA is 8.5 generations). But the likelihood
> of the common ancestor having lived precisely this generation is low
> (less than 4%).
>
> The 50% probability point is 21 generations (TMRCA is 10.5 gens).
> Half the time, the real MRCA will be within this period, and half the
> time outside of it.
>
> The 50% confidence interval is 14 to 29 generations (TMRCA is 7 to
> 14.5 generations).
>
> The 95% interval is 3 to 50 generations (TMRCA is 1.5 to 25
> generations).
>
> Identify your necessary confidence, then find your result. It's
> pretty easy.l
>
> You can add more markers to get narrower intervals.
>
> Vince

I second all Vince's comments and add that the intervals will get as narrow
as you like if you had unlimited STRs to add to the haplotypes --- it's
called the central limit theorem, but perhaps that's not part of "new math"?
But we don't have unlimited numbers of STRs; there are only so many useful
STRs in the ydna. At one point that number was reported on the list as only
something like 400 or so.
Ken


This thread: