Archiver > GENEALOGY-DNA > 2011-10 > 1318948232

From: "Kenneth Nordtvedt" <>
Subject: Re: [DNA] [yDNAhgI] New challenge8057 up
Date: Tue, 18 Oct 2011 08:30:32 -0600
References: <7D1794D27FA94D2983F248839CF8493F@kenPC><><><7C0D7A2EEA74413EBB2E56DA64CE65EF@kenPC><><><201A485353744DE79261E1D9117E1509@kenPC><><E6B738B5065440CBAD9C566A6BDF1050@kenPC><><26437F71DB8D437994F65A5CE8C1D1C8@kenPC><><B674EA6B21994D9C9047475B28982651@kenPC><>
In-Reply-To: <>

-----Original Message-----
From: Terry
Sent: Tuesday, October 18, 2011 12:16 AM
Subject: Re: [yDNAhgI] [DNA] New challenge8057 up

For 170 generations - do you mean 151 generations? You said 151 was the
answer, and the lockbox8057 file says G=151 for the genealogical age. Not
important, but 170.282 generations is the value you get from a TMRCA formula
if you knew the founder STR values - but we cannot assume we know those
founder STR values.

[[The purpose of the challenge files is to see how tree reconstruction
softwares reconstruct --- that includes in the ultimate and most ambitious
softwares, using final haplotypes and perhaps the str mutation rates,
estimating all the node positions (genealogy) and their ages, the haplotypes
at the nodes (this includes the founder haplotype), and maybe even the
specific mutations which happened on each tree branch segment. Not all
methods or softwares will be
doing all these things; but the lockbox is put out so people can have every
fact of the real tree and mutations for comparison after the try at

Real tmrca, tmrca by Var/M and coalescence age estimate are there for those
whose software may produce those things from their tree reconstruction
software instead of the variance methods, but may want to compare results
for those attributes. ]]

And for "28" (what you have called "pairs of octets"), I count 16 of
them. [[The 64 haplotypes consist of 8 octets, each descending from one of
eight nodes at a particular place upstream in the tree. You can pick 8 x 7
/ 2 = 28 pairs out of those 8 to get 16 descending haplotypes
Converting haplotype numbers back to the original ones using the code in
lockbox, the octets are haplotypes 1 to 8, 9 to 16, 17 to 24, ........ and
57 to 64. ]]

Finally you say: "But I can guarantee you the average variance tmrca age
over those 28 choices that could have been made, will be the same number
of generations as the overall haplotype collection tmrca estimate."

[[ The eight octets have tmrca estimates (var/m) of 186.7, 221.7, 211.7,
168.3, 166.7, 140, 1151.7, and 118.3 generations, respectively, for the
first eight, second eight, ........ last eight haplotypes in original
Average of those eight is exactly the 170.xy generations for the tmrca
estimate for all 64 haplotypes given in lockbox. ]]

Well let's see what you get then. The average I get (122.515) is a lot less
than 151.

[[You will have to give us the formula you use to produce the approximately
100 generations tmrca estimate for haplotypes 1 to 8 plus 57 to 64. Maybe
you are evaluating coalescence self-variance?

My tmrca estimate above are Sum over n [r(n)-rf]^2 with r(n) being repeat
value for some STR in haplotype n of some set of haplotypes, rf is the
founder repeat value; then summed over STRs, then that divided by sum of STR
mutation rates = .075 ]]

This thread: