World Families Forums - STR Wars: Is diversity meaningful? more meaningful than Hg frequency?

Welcome, Guest. Please login or register.
April 25, 2014, 03:47:31 AM
Home Help Search Login Register

+  World Families Forums
|-+  General Forums - Note: You must Be Logged In to post. Anyone can browse.
| |-+  R1b General (Moderator: rms2)
| | |-+  STR Wars: Is diversity meaningful? more meaningful than Hg frequency?
« previous next »
Pages: 1 ... 5 6 [7] 8 9 ... 14 Go Down Print
Author Topic: STR Wars: Is diversity meaningful? more meaningful than Hg frequency?  (Read 15981 times)
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #150 on: April 26, 2012, 12:56:19 PM »

Perhaps it would be good to know what methodology he used, because he gets a linearity that is three and four folds greater than the previously observed linearity based on the Busby et al(2011) study. ....  

I agree.  I do not understand how Marko T. Heinila derived the linear duration estimates.  I read his original postings at DNA forums (I think it was under U106 TMRCA report) and I was impressed with the large amount of data he used and his application of statistics, but he lost me. He is a scientist at the Helsinki University of Technology (something to do with their Low Temperature Lab.)

This is from memory, but my impression of Busby's research on linearity was that the data was very limited.  I've read on Rootsweb where you'd need to study some unbelievably high number of father-son transmissions to determine linearity that way.
 
Quote from: Busby
Fifteen Y-STRs with mutation rates, range of alleles and estimate of duration of linearity. All STRs investigated in this study are shown with their mutation rates (μ), estimated from Ballantyne et al. [33], and range of observed alleles, R, with 95% CI is taken from the YHRD [34]. θ(R)/2μ is an estimate of the duration of linearity of an STR."

In this an excerpt of that table with the duration of linearity column. I added the "S127"s and "xxx"s to show what markers he used in his S127(L11) STR variance calculations and which ones appeared to be too short in duration according to his own method to go back over 5000 years.
Quote from: Busby

Y-STR    θ(R)/2μ   
DYS448   25381   
DYS392   19244   S127
DYS438   12465   S127
DYS390   9211   S127
DYS393   5648   S127
DYS439   4861   S127 xxx
DYS437   4357   S127 xxx
DYS635    4221   
DYS456    3289   
DYS389II  3111   S127 xxx
DYS391    2554   S127 xxx
DYS458    1944   
DYS19      1888   S127 xxx
Y-GATA-H4 1630   
DYS389I    953   S127 xxx

Maybe you can figure this out.  This is where Busby lost me. Six of the ten STRs he used weren't fit close to for the purpose unless you assume R-L11 is even younger yet.  Why did he use these STRs in his analysis to show R-L11 was of about the same age all across Europe.? It's probably true, but it seems like he (Busby et al) argue against themselves in their efforts to knock Barlaresque.

Anyway, given 1) my inability to see consistency in Busby's stuff, 2) criticism that much large data sets on father-son events are needed, 3) Heinila's strong background and apparent grasp of statistics, 4) the large number of haplotypes Heinila analyzed, and 5) the fact that he analyzed all of FTDNA's 1st 67 STRs...     I started using Heinila's results as the basis for analyzing haplotype data with "linear" STRs only.

... but it doesn't matter much anyway, at least in the large R1b subclades I have data for.
« Last Edit: April 26, 2012, 01:14:36 PM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
JeanL
Old Hand
****
Offline Offline

Posts: 425


« Reply #151 on: April 26, 2012, 01:11:49 PM »

This is from memory, but my impression of Busby's research on linearity was that the data was very limited.  I've read on Rootsweb where you'd need to study some unbelievably high number of father-son transmissions to determine linearity that way.

So you think that calibration using some hypothetical clan project where everyone is presumed to be equally removed from the same ancestor, and that the mutation rates of STRs are linear throughout is a better fit, than father-son pairs where one can get a grasp of the actual number of mutations that occur across a generation?

The person who claims that father-son rates are useless is Klyosov, but we know that father-son rates are by far way more reliable than the mutation rates found using calibrations of the Donald Clan.

For once when one count mutations of a loci on a clan Project one is assuming the following:

1-Everyone is equally removed from the presumed common ancestor.
2- The haplotype that minimizes mutations is the modal haplotype, however we have no way of knowing if that was the haplotype of that ancestor, unless we actually test his DNA.
3-While loss of linearity might have no effect on determining the mutation of loci X in a time range of 1300 ybp, one can’t  extrapolate that longer time frames.
Logged
Jdean
Old Hand
****
Offline Offline

Posts: 678


« Reply #152 on: April 26, 2012, 01:19:41 PM »


I can't answer many of your queries.  I think it is important first to agree, or disagree, on my premise that many of the dys loci (medium rate) are limited/bounded.  I've provided a dataset that suggests they are, but I think we need more data.

A prior paper by goldstein, referenced in busby, gives a linearity equation.  Thats what busby used.  I don't know what range of values for each STR Markko used.  If he didn't recognize the problem with multisteps, I would question his definition of linearity.

I'm not asking you to run any test cases yet since I don't know how to specify what you are asking.  If someone who is much cleverer with S/W than I am could create some distribution tables, then we can evaluate that data and determine the next step.

I know Kens opinion of Zhiv.  That said, a lot of folks, as you know, who are knowledgeable are supportive of his approach.  What I'm trying to do is to come up with an understanding of why he had to fudge the data sets referenced by Chandler.  I don't think we are chasing ghosts here.

I appreciate all the attention you've paid to my comments.  I am limited in what guidance I can provide.

How big is this problem of multi step mutations ?

I can't say I come across it very often myself and just had a look at a large family group in a project run by a well known and respected researcher.

Out of the 67 people in this group there were a total of 90 off modal values in the first 37. Only three people had values that were greater than 1 from the modal values one was in YCAII and the other two in CDY and these loci aren't used in variance calculations.

70% of the loci in this group had at least one person with a mutation from the family modal.
« Last Edit: April 26, 2012, 01:44:31 PM by Jdean » Logged

Y-DNA R-DF49*
MtDNA J1c2e
Kit No. 117897
Ysearch 3BMC9

Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #153 on: April 26, 2012, 01:30:52 PM »

This is from memory, but my impression of Busby's research on linearity was that the data was very limited.  I've read on Rootsweb where you'd need to study some unbelievably high number of father-son transmissions to determine linearity that way.
So you think that calibration using some hypothetical clan project where everyone is presumed to be equally removed from the same ancestor, and that the mutation rates of STRs are linear throughout is a better fit, than father-son pairs where one can get a grasp of the actual number of mutations that occur across a generation?
No, no, I do not agree with Klyosov and in no way intend to support the hypothetical clan model. To me it would be flipping a coin ten times on the north pole and if came out 7 heads and 3 tails we could assume that to be the case everywhere millions of times over.

The person who claims that father-son rates are useless is Klyosov, but we know that father-son rates are by far way more reliable than the mutation rates found using calibrations of the Donald Clan.

For once when one count mutations of a loci on a clan Project one is assuming the following:

1-Everyone is equally removed from the presumed common ancestor.
2- The haplotype that minimizes mutations is the modal haplotype, however we have no way of knowing if that was the haplotype of that ancestor, unless we actually test his DNA.
3-While loss of linearity might have no effect on determining the mutation of loci X in a time range of 1300 ybp, one can’t  extrapolate that longer time frames.

I was talking about how Busby determined linearity timeframes (not mutation rates), but I honestly don't understand the "supposed" limitation in the data he use to calculate linearity.  I was going from memory on what I thought he did.  I'll go back and recheck. Maybe what he did is very thorough, it just seemed limited at the time I read it.

BTW, I did go back to the Myres and Barlaresque papers and reread where they got their data from and how they put it together.  I can repeat the variance calculations, etc., but I can't at all defend how they got their data from different sources.  It makes me nervous.
« Last Edit: April 26, 2012, 01:45:24 PM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #154 on: April 26, 2012, 01:42:58 PM »

How big is this problem of multi step mutations ?

I can't say I come across it very often myself and just had a look at a large family group in a project run by a well known and respected researcher.

Out of the 67 people in this group there were a total of 90 of modal values in the first 37. Only three people had values that were greater than 1 from the modal values one was in YCAII and the other two in CDY and these loci aren't used in variance calculations.

70% of the loci in this group had at least one person with a mutation from the family modal.

As far as I can tell it isn't a big problem, but there some people who are very concerned about it.

I do have one case in a project that I have where it appears that one subgroup jumped from 568=11 to 568=8, at least we think it did.  This does throw off TMRCA calculations when you have small groups.    When you examine the haplotypes, the faster STRs barely moved so we concluded this was a multi-step jump.

However, at the subclade level, R-L193 in this case, it gets washed out (TMRCA wise) anyway. It may be akin to saying one of the 67 STR experiments went bad, but with that many individual STR experiments and 100 haplotypes it doesn't impact anything.
« Last Edit: April 26, 2012, 01:44:45 PM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Jdean
Old Hand
****
Offline Offline

Posts: 678


« Reply #155 on: April 26, 2012, 01:56:20 PM »

How big is this problem of multi step mutations ?

I can't say I come across it very often myself and just had a look at a large family group in a project run by a well known and respected researcher.

Out of the 67 people in this group there were a total of 90 of modal values in the first 37. Only three people had values that were greater than 1 from the modal values one was in YCAII and the other two in CDY and these loci aren't used in variance calculations.

70% of the loci in this group had at least one person with a mutation from the family modal.

As far as I can tell it isn't a big problem, but there some people who are very concerned about it.

I do have one case in a project that I have where it appears that one subgroup jumped from 568=11 to 568=8, at least we think it did.  This does throw off TMRCA calculations when you have small groups.    When you examine the haplotypes, the faster STRs barely moved so we concluded this was a multi-step jump.

However, at the subclade level, R-L193 in this case, it gets washed out (TMRCA wise) anyway. It may be akin to saying one of the 67 STR experiments went bad, but with that many individual STR experiments and 100 haplotypes it doesn't impact anything.

My feelings exactly.

I recently came across somebody who had 12-30 at 389 in a relatively young cluster where as nearly everybody else had 12-28 which presumably was a jump of two, but it's not common.
Logged

Y-DNA R-DF49*
MtDNA J1c2e
Kit No. 117897
Ysearch 3BMC9

Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #156 on: April 26, 2012, 03:23:22 PM »

Perhaps it would be good to know what methodology he used, because he gets a linearity that is three and four folds greater than the previously observed linearity based on the Busby et al(2011) study. ....  

I agree.  I do not understand how Marko T. Heinila derived the linear duration estimates.  I read his original postings at DNA forums (I think it was under U106 TMRCA report) and I was impressed with the large amount of data he used and his application of statistics, but he lost me. He is a scientist at the Helsinki University of Technology (something to do with their Low Temperature Lab.)

He definitely analyzed the differences between up and down STR mutations. Unfortunately his links are no longer active.

Quote from: Rootsweb Heinila and Vernade
From: Marko Heinila
Subject: Re: [DNA] Understanding STR mutation rates
Date: Fri, 10 Jun 2011 09:37:32 -0600

Thanks for posting those links below. This is more obviously related
to the more detailed Markov chain model discussions that are
(finally!) becoming more popular on this list... the links provide
some information related the many parameters involved. There seems to
be no perfect estimation methods for these features, what is suggested
empirically in the links is hopefully better than nothing for a
starting point. Marko Heinila

From: vernade didier
Subject: Re: [DNA] Understanding STR mutation rates
Date: Tue, 7 Jun 2011 12:36:09 +0100 (BST)

This is off list because I don't know what Marko Heinila wants. I was
contacted by Marko Heinila and he provided links to his site :
http://beforepresent.dyndns.info/
http://beforepresent.dyndns.info/updownratio.php
http://beforepresent.dyndns.info/lengtheffect.php
http://beforepresent.dyndns.info/rates.php
http://beforepresent.dyndns.info/weights.php

The point is about mutation rates. This work seems interesting.
Logged

R1b-L21>L513(DF1)>L705.2
ironroad41
Old Hand
****
Offline Offline

Posts: 219


« Reply #157 on: April 26, 2012, 03:34:45 PM »

I have no theory on doggerland, its pure speculation from my point of view.

I very much respect Markos work.  I simply was trying to respond to your query. If Marko assumed that each dys locis spread was all due to single steps, then I would question how he defined linear based on work I did this winter.

As I understand it Marko is currently a quant working for a hedge fund in NYC and is now an american citizen.

Additionally, I have no agenda.  I simply think I may have an idea why the Zhiv folks had to apply a fudge factor to father/son rates.

As I pointed out previously, we need to better understand all aspects of the Y STR mutational process based on the data that has been accumulated.

re: multisteps.  I have seen estimates as high as 10% of all mutations, but I think 5% is a more solid number.  A multistep of 4 generates 16 times as much variance/ASD as a single step at a dys loci, thats just the way the model works.
« Last Edit: April 26, 2012, 03:35:17 PM by ironroad41 » Logged
razyn
Old Hand
****
Offline Offline

Posts: 405


« Reply #158 on: April 26, 2012, 04:19:43 PM »

Perhaps it would be good to know what methodology he used, because he gets a linearity that is three and four folds greater than the previously observed linearity based on the Busby et al(2011) study. ....  

I agree.  I do not understand how Marko T. Heinila derived the linear duration estimates.  I read his original postings at DNA forums (I think it was under U106 TMRCA report) and I was impressed with the large amount of data he used and his application of statistics, but he lost me. He is a scientist at the Helsinki University of Technology (something to do with their Low Temperature Lab.)

He definitely analyzed the differences between up and down STR mutations. Unfortunately his links are no longer active.

His links aren't, but he is.  A few days ago, he was recovering from long distance flights.  He's aware of the current activity on this forum, and with luck we may hear from him.  He will be analyzing data from 4/17/12, which is recent enough to have results for several of the new SNPs under Z196 (among other new things).  I don't know when his new analysis will be complete, what length haplotypes it will cover, or where it will be visible.  But I'm sure it will be interesting, if we can find it.
Logged

R1b Z196*
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #159 on: April 27, 2012, 09:54:57 AM »

I have no theory on doggerland, its pure speculation from my point of view.

I very much respect Markos work.  I simply was trying to respond to your query. If Marko assumed that each dys locis spread was all due to single steps, then I would question how he defined linear based on work I did this winter.

As I understand it Marko is currently a quant working for a hedge fund in NYC and is now an american citizen.

Additionally, I have no agenda.  I simply think I may have an idea why the Zhiv folks had to apply a fudge factor to father/son rates.

As I pointed out previously, we need to better understand all aspects of the Y STR mutational process based on the data that has been accumulated.

re: multisteps.  I have seen estimates as high as 10% of all mutations, but I think 5% is a more solid number.  A multistep of 4 generates 16 times as much variance/ASD as a single step at a dys loci, thats just the way the model works.

Its okay to have questions and challenges, but it also important to do some homework. We all had access to his web links last year. I thought I'd better get the information while I could so he sent me a file folder of the documentation he had there when he closed it down.

I don't know if Marko's model is correct, but he certainly made efforts to account for concerns like the ones you have.  As you can see, he tries to calculate both up and down rates as well as multi-steps.

Quote from: Marko Heinila web link 2011

Y-Tree Mutation Rates

locus    mean RN    population rate    modal RN    modal rate    modal downrate    modal uprate    multisteps    Delta ln(uprate)    Delta ln(downrate)
DYS393    13.1    0.00104    13    0.00095    0.000336    0.000611    0.03    0.11    0.94
DYS390    23.6    0.00228    24    0.00243    0.001202    0.001227    0.02    0.15    0.58
DYS19a    14.3    0.00161    14    0.00129    0.000241    0.001046    0.03    0.23    0.96
DYS391    10.5    0.00289    10    0.00149    0.000355    0.001137    0.01    0.00    1.88
DYS385a    12.1    0.00172    11    0.00134    0.000421    0.000921    0.08    0.11    0.29
DYS385b    14.6    0.00334    14    0.00294    0.001112    0.001829    0.10    0.11    0.22
DYS426    11.6    0.00013    12    0.00017    0.000081    0.000090    0.07    0.95    0.82
DYS388    12.6    0.00052    12    0.00034    0.000074    0.000263    0.15    0.33    0.57
DYS439    11.6    0.00499    12    0.00523    0.002273    0.002953    0.00    0.25    1.02
DYS389i    13.0    0.00213    13    0.00188    0.000615    0.001267    0.02    0.05    1.26
DYS392    12.3    0.00058    13    0.00069    0.000221    0.000469    0.10    0.34    0.51
DYS389B    16.4    0.00292    16    0.00231    0.000624    0.001689    0.03    0.18    0.86
DYS458    16.7    0.00697    17    0.00711    0.003145    0.003968    0.04    0.17    0.40
DYS459a    8.7    0.00045    9    0.00043    0.000286    0.000141    0.07    -0.85    1.19
DYS459b    9.6    0.00118    10    0.00152    0.001339    0.000177    0.01    -0.81    1.87
DYS455    10.6    0.00026    11    0.00028    0.000151    0.000130    0.06    0.17    0.46
DYS454    11.1    0.00029    11    0.00026    0.000061    0.000198    0.09    0.52    1.40
DYS447    24.7    0.00304    25    0.00313    0.001629    0.001501    0.06    0.11    0.12
DYS437    14.9    0.00085    15    0.00082    0.000509    0.000314    0.03    0.08    0.74
DYS448    19.4    0.00138    19    0.00124    0.000593    0.000647    0.03    0.14    0.31
DYS449    29.5    0.00843    29    0.00777    0.003120    0.004647    0.05    0.06    0.18
DYS460    10.7    0.00342    11    0.00346    0.001811    0.001652    0.01    -0.09    1.42
Y-GATA-H4    10.7    0.00230    11    0.00248    0.001398    0.001082    0.01    0.12    0.91
YCAIIa    19.2    0.00046    19    0.00042    0.000188    0.000228    0.30    -0.04    0.51
YCAIIb    22.1    0.00094    23    0.00106    0.000737    0.000326    0.18    -0.03    0.27
DYS456    15.4    0.00508    15    0.00380    0.001457    0.002340    0.02    0.17    0.70
DYS607    14.6    0.00217    15    0.00235    0.001288    0.001062    0.04    0.18    0.38
DYS576    17.5    0.01085    18    0.01152    0.005357    0.006166    0.03    0.14    0.44
DYS570    17.9    0.00874    17    0.00677    0.002876    0.003894    0.04    0.17    0.28
CDYa    35.4    0.01265    36    0.01300    0.006602    0.006399    0.07    0.02    0.14
CDYb    37.5    0.01632    38    0.01681    0.008045    0.008763    0.08    0.08    0.18
DYS442    12.0    0.00317    12    0.00291    0.001211    0.001699    0.02    0.24    0.73
DYS438    11.1    0.00049    12    0.00058    0.000270    0.000305    0.07    0.08    0.37
DYS531    11.0    0.00044    11    0.00043    0.000088    0.000339    0.08    0.34    1.05
DYS578    8.5    0.00019    9    0.00023    0.000064    0.000163    0.08    0.32    1.08
DYF395S1a    15.2    0.00042    15    0.00040    0.000113    0.000286    0.08    -0.66    0.78
DYF395S1b    15.9    0.00032    16    0.00031    0.000143    0.000171    0.07    0.38    0.64
DYS590    8.0    0.00013    8    0.00013    0.000045    0.000084    0.07    
   
DYS537    10.5    0.00104    10    0.00076    0.000088    0.000671    0.00    0.03    1.49
DYS641    10.0    0.00030    10    0.00030    0.000066    0.000238    0.05    
   
DYS472    8.0    0.00002    8    0.00002    0.000000    0.000016    0.00    
   
DYF406S1    10.3    0.00164    10    0.00137    0.000422    0.000950    0.02    0.19    0.67
DYS511    9.9    0.00138    10    0.00133    0.000456    0.000877    0.02    0.19    1.22
DYS425    12.1    0.00013    12    0.00012    0.000059    0.000064    0.19    0.53    0.49
DYS413a    22.0    0.00197    23    0.00233    0.001778    0.000552    0.20    -0.06    0.32
DYS413b    22.8    0.00139    23    0.00136    0.000579    0.000781    0.17    0.08    0.46
DYS557    15.9    0.00331    16    0.00327    0.001308    0.001958    0.05    0.12    0.27
DYS594    10.2    0.00046    10    0.00044    0.000109    0.000331    0.08    -0.04    0.67
DYS436    12.0    0.00010    12    0.00010    0.000029    0.000070    0.18    
   
DYS490    12.0    0.00021    12    0.00021    0.000041    0.000167    0.19    0.36    0.64
DYS534    15.4    0.00755    15    0.00643    0.002552    0.003879    0.04    0.14    0.41
DYS450    8.0    0.00018    8    0.00018    0.000055    0.000125    0.10    0.33    0.54
DYS444    12.4    0.00334    12    0.00269    0.000993    0.001700    0.01    0.14    0.68
DYS481    23.1    0.00421    22    0.00334    0.001164    0.002178    0.11    0.10    0.25
DYS520    20.2    0.00171    20    0.00159    0.000565    0.001026    0.03    0.14    0.32
DYS446    13.2    0.00335    13    0.00312    0.001210    0.001906    0.04    0.13    0.23
DYS617    12.3    0.00061    12    0.00054    0.000150    0.000389    0.08    0.31    0.44
DYS568    11.1    0.00048    11    0.00040    0.000181    0.000224    0.08    0.26    1.41
DYS487    13.0    0.00087    13    0.00081    0.000172    0.000638    0.10    0.23    1.02
DYS572    10.9    0.00113    11    0.00110    0.000733    0.000372    0.02    -0.23    1.06
DYS640    11.2    0.00037    11    0.00029    0.000039    0.000256    0.04    0.13    2.12
DYS492    12.1    0.00025    12    0.00023    0.000068    0.000163    0.11    0.34    0.64
DYS565    11.7    0.00073    12    0.00081    0.000502    0.000309    0.05    0.29    1.26

Once again, let me reiterate that the fine tuning and disagreement among hobbyist-scientists doesn't matter much, at least in terms of R1b.

Heinila's intraclade TMRCA for R-L21 is 4.2k ybp. His interclade for the L21-U152 TMRCA is also 4.2k ybp.  Anatole Klosov gets similar estimates for R-L21. I don't know if he does interclade estimates.  Using Ken Nordtvedt's Generation 7, I get the intraclade Coalescence Age for R-L21 as 3.7k ybp. With Gen 7, I get the L21-U152 interclade TMRCA as 4.4k ybp. The R1b subclade TMRCA estimates line up nicely across the board for these guys.

Quote from: Marko Heinila web link 2011

Time Values

haplogroup ____ shorthand _ proven _ predicted  intraclade interclade _ interclade_hg...
_________________________ members _ members (kyBP) (kyBP)
R1b1a2a1a1a _____ R-U106 ___ 1203 __ 3720 __ 4.5 ___ 4.5 __ R1b1a2a1a1b R-P312
R1b1a2a1a1b _____ R-P312 ___ 3403 __ 10724 _ 4.3 ___ 4.5 __ R1b1a2a1a1a R-U106
R1b1a2a1a1b3 ____ R-U152 ___ 439 ___ 1502 __ 4.2 ___ 4.2 __ R1b1a2a1a1b4 R-L21
R1b1a2a1a1b3c ___ R-L2 _____ 254 ___ 996 ___ 4.2 ___ 4.2 __ R1b1a2a1a1b4 R-L21
R1b1a2a1a1b4 ____ R-L21 ____ 2038 __ 6813 __ 4.2 ___ 4.2 __ R1b1a2a1a1b3  R-U152
R1b1a2a1a1b5a ___ R-SRY2627_ 126 ___ 341 ___ 3.5 ___ 3.8 __ R1b1a2a1a1b5b    R-L165

Moving up a step on the Y tree, you can see that Marko has the interclade TMRCA for P312-U106 as 4.5k ybp. Using Ken's method I get 4.5k ybp as well.  I gather Maliclavelli thinks it is ridiculous but I think you can see why I think these major R-L11 subclades were essentially of one clan or tribe at one time.
« Last Edit: April 27, 2012, 10:27:41 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #160 on: April 27, 2012, 11:22:52 AM »

@Mikewww or anyone familiar with Generations7 spreadsheet

Do you know if there is an explanation somewhere as to what math operation Ken uses to account for hidden mutations?  

You should probably look at his formulas and his powerpoint charts where he charts out and tries to explain his methodology.

I think the answer is something along the lines of what John Chandler is saying.  This is from reply #5 of this thread.

A recent conversation from Rootsweb:
Quote from: general question
My own layman's viewpoint has always been to wonder how such unknowable factors like bottle-necks, back mutations, etc. can ever be adequately compensated for
Here is a response from a Scientist at MIT. John Chandler is the guy who calculated the mutation rates most of us use.
Quote from: John Chandler
That "etc." is exactly the difficulty. I'll point out in passing that back mutations are automatically accounted for in the variance method, ...
http://archiver.rootsweb.ancestry.com/th/read/genealogy-dna/2012-03/1333051203

My understanding of the explanation is that their mathematical model does not care about hidden mutations or even multi-step mutations. The mutation rates were derived based on visible mutations so, as long as they have adequate data to build the mutation rates, the way the TMRCA method uses them is consistent.  We should not think of the published mutation rate as literally the physical rate of change per the STR, but rather the observable rate of change.

What is required is that the STRs act somewhat consistently, in other words the expected (predicted) rates up and down should be the same and the rates shouldn't change given the allele value, etc.   This would be where the concern about STRs reaching saturation and high alleles values comes into play.  If an STR doesn't show linear duration (of its rate) during the timeframe we care about then it is not helpful.   The goal of the math model is to include STRs that are linear or "on average" (in aggregate) linear.

Today, Ken Nordtvedt posted a response related to some of these questions.
Quote from: Ken Nordtvedt
After all these years folks on (other) forums are still wondering or doubting variance measure between haplotypes properly counts back mutations. It does

Suppose an STR mutates twice between two haplotypes. The two could be up,up or up,down or down,up, or down,down. These four possibilities have equal probabilities of 1/4 if up and down mutations are equally likely.

So expected value of variance (which involves squaring STR repeat differences) is 4 x 2/4 + 0 x 2/4 = 2

Suppose an STR mutates three times between two haplotypes. The three could be up,up,up or up,up,down or up,down,up or down,up,up, or up,down,down, or down,up,down, or down,down,up, or down,down,down
Each of these eight alternatives of three mutations has probability 1/8.

So expected value of variance (which involves squaring STR repeat differences) is 9 x 2/8 + 1 x 6/8 = 3

Etc. Yes, variance totally takes into account back mutations and multiple mutations in same direction.......
http://archiver.rootsweb.ancestry.com/th/read/Y-DNA-HAPLOGROUP-I/2012-04/1335537421
« Last Edit: April 27, 2012, 11:23:31 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
JeanL
Old Hand
****
Offline Offline

Posts: 425


« Reply #161 on: April 27, 2012, 11:32:52 AM »


Today, Ken Nordtvedt posted a response related to some of these questions.
Quote from: Ken Nordtvedt
After all these years folks on (other) forums are still wondering or doubting variance measure between haplotypes properly counts back mutations. It does

Suppose an STR mutates twice between two haplotypes. The two could be up,up or up,down or down,up, or down,down. These four possibilities have equal probabilities of 1/4 if up and down mutations are equally likely.

So expected value of variance (which involves squaring STR repeat differences) is 4 x 2/4 + 0 x 2/4 = 2

Suppose an STR mutates three times between two haplotypes. The three could be up,up,up or up,up,down or up,down,up or down,up,up, or up,down,down, or down,up,down, or down,down,up, or down,down,down
Each of these eight alternatives of three mutations has probability 1/8.

So expected value of variance (which involves squaring STR repeat differences) is 9 x 2/8 + 1 x 6/8 = 3

Etc. Yes, variance totally takes into account back mutations and multiple mutations in same direction.......
http://archiver.rootsweb.ancestry.com/th/read/Y-DNA-HAPLOGROUP-I/2012-04/1335537421


What I bolded is a key assumption, that up and down mutations are equally likely(i.e. That they have the same mutation rates) to occur, in reality, they aren't, they have different mutation rates.
« Last Edit: April 27, 2012, 11:33:18 AM by JeanL » Logged
Jdean
Old Hand
****
Offline Offline

Posts: 678


« Reply #162 on: April 27, 2012, 12:43:27 PM »


Today, Ken Nordtvedt posted a response related to some of these questions.
Quote from: Ken Nordtvedt
After all these years folks on (other) forums are still wondering or doubting variance measure between haplotypes properly counts back mutations. It does

Suppose an STR mutates twice between two haplotypes. The two could be up,up or up,down or down,up, or down,down. These four possibilities have equal probabilities of 1/4 if up and down mutations are equally likely.

So expected value of variance (which involves squaring STR repeat differences) is 4 x 2/4 + 0 x 2/4 = 2

Suppose an STR mutates three times between two haplotypes. The three could be up,up,up or up,up,down or up,down,up or down,up,up, or up,down,down, or down,up,down, or down,down,up, or down,down,down
Each of these eight alternatives of three mutations has probability 1/8.

So expected value of variance (which involves squaring STR repeat differences) is 9 x 2/8 + 1 x 6/8 = 3

Etc. Yes, variance totally takes into account back mutations and multiple mutations in same direction.......
http://archiver.rootsweb.ancestry.com/th/read/Y-DNA-HAPLOGROUP-I/2012-04/1335537421


What I bolded is a key assumption, that up and down mutations are equally likely(i.e. That they have the same mutation rates) to occur, in reality, they aren't, they have different mutation rates.

But are they different enough to cause the calculations to be out by much, and if so how difficult would it be to correct ?
Logged

Y-DNA R-DF49*
MtDNA J1c2e
Kit No. 117897
Ysearch 3BMC9

Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #163 on: April 27, 2012, 01:03:12 PM »


Today, Ken Nordtvedt posted a response related to some of these questions.
Quote from: Ken Nordtvedt
After all these years folks on (other) forums are still wondering or doubting variance measure between haplotypes properly counts back mutations. It does

Suppose an STR mutates twice between two haplotypes. The two could be up,up or up,down or down,up, or down,down. These four possibilities have equal probabilities of 1/4 if up and down mutations are equally likely.

So expected value of variance (which involves squaring STR repeat differences) is 4 x 2/4 + 0 x 2/4 = 2

Suppose an STR mutates three times between two haplotypes. The three could be up,up,up or up,up,down or up,down,up or down,up,up, or up,down,down, or down,up,down, or down,down,up, or down,down,down
Each of these eight alternatives of three mutations has probability 1/8.

So expected value of variance (which involves squaring STR repeat differences) is 9 x 2/8 + 1 x 6/8 = 3

Etc. Yes, variance totally takes into account back mutations and multiple mutations in same direction.......
http://archiver.rootsweb.ancestry.com/th/read/Y-DNA-HAPLOGROUP-I/2012-04/1335537421


What I bolded is a key assumption, that up and down mutations are equally likely(i.e. That they have the same mutation rates) to occur, in reality, they aren't, they have different mutation rates.

But are they different enough to cause the calculations to be out by much, and if so how difficult would it be to correct ?

Ken will respond quickly today, I think, if you ask your questions.  
To subscribe to that forum, please send an email to Y-DNA-HAPLOGROUP-I-request@rootsweb.com with the word 'subscribe' without the quotes in the subject and the body of the message.

I'm sure he has thought about your issues and run simulations to determine how he should handle via his methodologies, but if you are uncomfortable that he either hasn't worked this out thoroughly or that he doesn't understand then you should ask him directly. My only advise is to read his web site documentation thoroughly first so we don't waste his time.

My opinion is that Nordtvedt really understands and I know he runs simulations against different mathematically modeled concepts.  I'm not saying you think differently of him, but you should delve into the areas you are concerned about directly.

I've got the conversation going on this topic at the forum Ken frequents, so please jump in.

I think it is easy to misunderstand simple illustrations. One of your concerns is his assumptions that
Quote from: JeanL
that up and down mutations are equally likely(i.e. That they have the same mutation rates) to occur, in reality, they aren't, they have different mutation rates.

I think he is just trying to simply illustrate the concept he is using so that a lay person could understand it.  If you get deeper into these assumptions I think you will find that he has thoroughly investigated errors or aberrant data to determine how to control and report confidence intervals and improve precision.  In other words, you'll quickly dive deep into some math formulas and the understanding of the terms and conditions in the equations will be the terminology of the conversation.

However, please, please, please, do not take my word on any of this. Go to Rootsweb Hg I and ask Ken what you wish.
« Last Edit: April 27, 2012, 01:10:54 PM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #164 on: April 27, 2012, 01:17:01 PM »

@Mikewww or anyone familiar with Generations7 spreadsheet

Do you know if there is an explanation somewhere as to what math operation Ken uses to account for hidden mutations?  

You should probably look at his formulas and his powerpoint charts where he charts out and tries to explain his methodology.

I think the answer is something along the lines of what John Chandler is saying.  This is from reply #5 of this thread.

A recent conversation from Rootsweb:
Quote from: general question
My own layman's viewpoint has always been to wonder how such unknowable factors like bottle-necks, back mutations, etc. can ever be adequately compensated for
Here is a response from a Scientist at MIT. John Chandler is the guy who calculated the mutation rates most of us use.
Quote from: John Chandler
That "etc." is exactly the difficulty. I'll point out in passing that back mutations are automatically accounted for in the variance method, ...
http://archiver.rootsweb.ancestry.com/th/read/genealogy-dna/2012-03/1333051203

My understanding of the explanation is that their mathematical model does not care about hidden mutations or even multi-step mutations. The mutation rates were derived based on visible mutations so, as long as they have adequate data to build the mutation rates, the way the TMRCA method uses them is consistent.  We should not think of the published mutation rate as literally the physical rate of change per the STR, but rather the observable rate of change.

What is required is that the STRs act somewhat consistently, in other words the expected (predicted) rates up and down should be the same and the rates shouldn't change given the allele value, etc.   This would be where the concern about STRs reaching saturation and high alleles values comes into play.  If an STR doesn't show linear duration (of its rate) during the timeframe we care about then it is not helpful.   The goal of the math model is to include STRs that are linear or "on average" (in aggregate) linear.

Today, Ken Nordtvedt posted a response related to some of these questions.
Quote from: Ken Nordtvedt
After all these years folks on (other) forums are still wondering or doubting variance measure between haplotypes properly counts back mutations. It does

Suppose an STR mutates twice between two haplotypes. The two could be up,up or up,down or down,up, or down,down. These four possibilities have equal probabilities of 1/4 if up and down mutations are equally likely.

So expected value of variance (which involves squaring STR repeat differences) is 4 x 2/4 + 0 x 2/4 = 2

Suppose an STR mutates three times between two haplotypes. The three could be up,up,up or up,up,down or up,down,up or down,up,up, or up,down,down, or down,up,down, or down,down,up, or down,down,down
Each of these eight alternatives of three mutations has probability 1/8.

So expected value of variance (which involves squaring STR repeat differences) is 9 x 2/8 + 1 x 6/8 = 3

Etc. Yes, variance totally takes into account back mutations and multiple mutations in same direction.......
http://archiver.rootsweb.ancestry.com/th/read/Y-DNA-HAPLOGROUP-I/2012-04/1335537421

I continued the conversation and asked him - As far as multi-increment single event mutations, what is the best way to handle them?  In one of my projects I edit what appears to be an
obvious jump before calculating TRMCAs.  In large haplogroup related calculations, these seem to "wash out" so I don't worry about them. Are there recommendations or guidelines for them?

Quote from: Ken Nordtvedt
Problem with multi-step mutations --- and we do know they happen now and then --- is that we do not know their rates very well. Actually, we don't know individual STR single step mutation rates too well.

But if n step mutation rates are known, equal up and down (another fiction), then they are easily included in mutation rate and contribute to variance in a straightforward manner by changing:

m --> m* = m(1) + 4 m(2) + 9 m(3) + .... with m(n) being mutation rate for n step mutations. In other words, don't worry about multiple steps and just evaluate variance according to normal rules, but adjust the STR mutation rates upward according to formula given.

Of course the higher step mutations wreck havoc with statistical sigma for variance after G generations --- sigmas go up significantly.

Note: m ---> m* is necessarily an INCREASE, which DECREASES tmrca estimates.

I also asked - Is the "maximum likelihood" method useful for TMRCA calculations? Does it have advantages or disadvantages? I've looked at Marko Heinila's work on this and his TMRCA estimates for R1b subclades come out very similar to what I get when I use your Generations 7 methodology.

Quote from: Ken Nordtvedt
For branch segments which are not too long, GD and GD-based maximum liklihood should not differ too much from variance as long as many STRs are used.
But there are devils in the details of programs which look for maximum liklihoods of trees; so one should know what's in the black boxes.

I'm not saying Nordtvedt's methodologies are perfect, nor even perfectly precise. They aren't, but at least he takes great care in understanding and testing issues and managing (and reporting) the confidence intervals. His interclade methodology is probably the best thing out there... just my opinion.

Did you notice how he note that the differences in the methodologies shouldn't make much difference if a lot of STRs are used?   That's what I've been trying to say.   If we use a lot STRs and a lot of haplotypes, the results don't vary much through various iterations.
« Last Edit: April 27, 2012, 05:24:53 PM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #165 on: April 27, 2012, 01:19:02 PM »

For the Archive access,

http://archiver.rootsweb.ancestry.com/th/index/Y-DNA-HAPLOGROUP-I

Logged

148326
Pos: Z245 L459 L21 DF13**
Neg: DF23 L513 L96 L144 Z255 Z253 DF21 DF41 (Z254 P66 P314.2 M37 M222  L563 L526 L226 L195 L193 L192.1 L159.2 L130 DF63 DF5 DF49)
WTYNeg: L555 L371 (L9/L10 L370 L302/L319.1 L554 L564 L577 P69 L626 L627 L643 L679)
JeanL
Old Hand
****
Offline Offline

Posts: 425


« Reply #166 on: April 27, 2012, 05:48:51 PM »


My opinion is that Nordtvedt really understands and I know he runs simulations against different mathematically modeled concepts.  I'm not saying you think differently of him, but you should delve into the areas you are concerned about directly.


The problem is that in a simulation everything will turn out as expected, which is why we collect empirical data to compare against the simulations. Therefore the simulations are used as the benchmark, the ideal that should be, but often times the reality is quite different from the ideal.

There was a recently(2011) published paper called:

Germline mutations of STR-alleles include multi-step mutations as defined by sequencing of repeat and flanking regions

Quote from: Dauber et al(2011)
Well defined estimates of mutation rates are a prerequisite for the use of short tandem repeat (STR-) loci in relationship testing. We investigated 65 isolated genetic inconsistencies, which were observed within 50,796 allelic transfers at 23 STR-loci (ACTBP2 (SE33), CD4, CSF1PO, F13A1, F13B, FES, FGA, vWA, TH01, TPOX, D2S1338, D3S1358, D5S818, D7S820, D8S1132, D8S1179, D12S391, D13S317, D16S539, D17S976, D18S51, D19S433, D21S11) in Caucasoid families residing in Austria and Switzerland. Sequencing data of repeat and flanking regions and the median of all theoretically possible mutational steps showed valuable information to characterise the mutational events with regard to parental origin, change of repeat number (mutational step size) and direction of mutation (losses and gains of repeats). Apart from predominant single-step mutations including one case with a double genetic inconsistency, two double-step and two apparent four-stepmutations could be identified. More losses than gains of repeats and more mutations originating from the paternal than the maternal lineage were observed (31 losses, 22 gains, 12 losses or gains and 47 paternal, 11 maternal mutations and 7 unclear of parental origin). The mutation in the paternal germline was 3.3 times higher than in the maternal germline. The results of our study show, that apart from the vast majority of single-step mutations rare multi-step mutations can be observed. Therefore, the interpretation of mutational events should not rigidly be restricted to the shortest possible mutational step, because rare but true multi-stepmutations can easily be overlooked, if haplotype analysis is not possible.

Logged
ironroad41
Old Hand
****
Offline Offline

Posts: 219


« Reply #167 on: April 28, 2012, 08:42:53 AM »

I have analyzed the data set I referenced earlier and it is clear to me that up/down rates are not equal.  As Razyn points out though, How important is that issue?

I find that several of Kens assumptions are not met:  equal up down rates and equally likely forward and backward mutations.  For a tightly knit dys loci, as many are, If you are at +1 from the modal, there is little or no probability of mutating to step two.  The only possible step is then back to the modal.

By definition of modal, it is the value most frequently observed, we have no way I can currently think of to count the number of mutations that have occurred from the modal and then back.  We really can't depend on the count on the number of mutations which have occurred in the past at any allele value since we don't know the time history of the mutational process.

I am starting the slow process (for me) of creating a distribution table for as many of the R- L21 dys loci as I can.  It is already pretty obvious that the up/down rates are not similar, as previous data suggests.  The bigger question is the issue of bounded mutations at a dys loci and the impact of multisteps.  For the faster mutators such as CDYa,b it will probably be impossible to sort out single steps from multisteps.  From distributions I have observed in the past the faster mutators approach a distribution which is almost uniform, two to four alleles with similar values and then rapidly falling off tails.  It seems obvious to me that all dys loci have a set of bounded values which is exacerbated by the occasional multistep.

I reiterate my objective as being to determine if we can explain the "fudge factor" of Zhivotovsky from properties of the STR mutational process.  Something is changing the father/son meioses data as the real process plays out.  JMHO.
Logged
Maliclavelli
Guru
*****
Online Online

Posts: 2115


« Reply #168 on: April 28, 2012, 10:49:23 AM »

I find that several of Kens assumptions are not met:  equal up down rates and equally likely forward and backward mutations.  For a tightly knit dys loci, as many are, If you are at +1 from the modal, there is little or no probability of mutating to step two.  The only possible step is then back to the modal.

By definition of modal, it is the value most frequently observed, we have no way I can currently think of to count the number of mutations that have occurred from the modal and then back.  We really can't depend on the count on the number of mutations which have occurred in the past at any allele value since we don't know the time history of the mutational process.
I thank you for what you are saying. I have expressed these concepts in the past with my golden principles:
1) mutations happen around the modal
2) there is a convergence to the modal as time passes
3) sometime a mutation goes for the tangent (and we have the outlier)

This said, my analysis of calculating the time passed from the BT haplogroup by counting how the modal of BT has generated other values forwards and backwards, done all the due calculations, perhaps merits to be taken in consideration.
Logged

Maliclavelli


YDNA: R-S12460


MtDNA: K1a1b1e

Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #169 on: April 28, 2012, 10:50:21 AM »

I have analyzed the data set I referenced earlier and it is clear to me that up/down rates are not equal.  As Razyn points out though, How important is that issue?

I find that several of Kens assumptions are not met:  equal up down rates and equally likely forward and backward mutations.  

Keep in mind that Marko Heinila accounts for up and down mutation rates as well as multi-increment STR mutations and comes out with about the same TMRCA estimations as Nordtvedt's - as I've shown you earlier in this thread.

.... I reiterate my objective as being to determine if we can explain the "fudge factor" of Zhivotovsky from properties of the STR mutational process.  Something is changing the father/son meioses data as the real process plays out.  JMHO.
I'm sorry, I didn't realize that was your objective.

If it is Zhivotosky's "fudge factor" then why should we attempt to explain it? You should consider reviewing his studies closely and determining if you think his conclusions are valid.  If he can't explain to your satisfaction then that it is Zhivotosky who you should be questioning.  

Do you think his rates are the correct ones? If so, why?

In terms of the relative STR diversity between known groups of related people (i.e. haplogroups), the mutation rate doesn't really matter. We aren't applying time to the result when just looking at variance. Their relative positioning amongst themselves (the haplogroups) can be useful information but you can slide the time scale forward or backward as you wish as different assumptions.   Just keep in mind that if you slide the timeline back (slower mutation rates) then you have to account for the archaeology of Homo Sapiens Sapiens' (modern man) appearance in Europe.
Quote from: Wikipedia
H. sapiens reached Europe around 40,000 years ago
« Last Edit: April 28, 2012, 10:58:32 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Maliclavelli
Guru
*****
Online Online

Posts: 2115


« Reply #170 on: April 28, 2012, 10:58:46 AM »

The analyses of aDNA in many places of Europe have demonstrated that the TMRCA of that haplotypes, calculated by your method, was at least less for a 2.5 factor.
Voilà Zhivotovsky!
Logged

Maliclavelli


YDNA: R-S12460


MtDNA: K1a1b1e

Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #171 on: April 28, 2012, 11:02:16 AM »

My opinion is that Nordtvedt really understands and I know he runs simulations against different mathematically modeled concepts.  I'm not saying you think differently of him, but you should delve into the areas you are concerned about directly.

The problem is that in a simulation everything will turn out as expected, which is why we collect empirical data to compare against the simulations. Therefore the simulations are used as the benchmark, the ideal that should be, but often times the reality is quite different from the ideal.

Nothing is every ideal, which is what science tries to deal with. Everything I've read from Nordtvedt is that he is evaluating the data and trying to account for what he calls "fictions." That doesn't mean the mathematical models don't work.. it's a matter of precision.

I have no issue with your concern. I don't think your concern has a significant impact given the number of markers I'm evaluating. As I've shown on this thread, the results seem to show a correlation with the known SNP defined phylogenetic Y DNA tree, at least for the R1b subclade.  We know Ken thinks the correlation of STR diversity to time applies to the Hg I subclades appropriately as you can see from his web site.

... but I am NOT an expert so I'll request again:

Please, please, please make your challenges directly to Nordtvedt on the Rootsweb haplogroup I forum. I'm trying to be a proxy for some of your concerns, but you are much better at arguing your point of view than I am so please go ask Ken.
« Last Edit: April 28, 2012, 11:17:18 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #172 on: April 28, 2012, 11:10:54 AM »

The analyses of aDNA in many places of Europe have demonstrated that the TMRCA of that haplotypes, calculated by your method, was at least less for a 2.5 factor.
Voilà Zhivotovsky!
Please be specific on how that is demonstrated.  What level of SNP testing for what sample and in comparison to what hobbyist-scientist's TMRCA estimate? Please specifically direct us to the TMRCA estimate that you are challenging. I'm not well versed in non-R1b SNPs so please provide as much relevant background as you can on the other haplogroups if that is your area of challenge.

BTW, I think you realize that the TMRCA for a group people living today is not necessarily the same thing as the original bearer or an early bearer of an SNP.  The surviving folks' MRCA may have come from a totally different place.

I agree with you that there is a chance that the MRCA may only represent one lopsided branch of an old haplogroup. There is no doubt that this is a real possibility.  What is the question we seek to answer? ...
Is it when was an SNP born or is when was the TMRCA for a group of people living today and where did they come from?
« Last Edit: April 28, 2012, 11:30:30 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
ironroad41
Old Hand
****
Offline Offline

Posts: 219


« Reply #173 on: April 28, 2012, 11:19:38 AM »

As I remember markos work he also had an unexplained 5K to 6K years where he couldn't find enough data between L11 and subsequent mutations?

I have read Zhiv's papers and understand them as well as I can.  He doesn't try to explain the difference in the first paper and has been on a hunt in subsequent papers.  At this point in time, I don't think he knows why either.  An additional point is that he may have not recognized the issues of unique event mutations and multisteps and this omission may be part of his fudge factor for all I know.

I happen to believe it is in the intrinsic mutation process and the models being used to represent the process.  Up/down rates are generally not the same, Many loci are bounded around the modal, these are both observable data and do not match the criterion Ken put forth.  Marko did try to include up/down rates, I'm not sure about multisteps and I don't think he considered/evalueated the impact on his model of bounded dys loci.

I think that Zhivs estimates are better applied to larger time estimates than 1K or so.  Shorter in time than that I can show that father/son rates are directly applicable.  I've applied this technique to both Kerchners family and the Ian Cam and other clans and got sensible results. In both cases the probability of a hidden mutation at the medium to slow dys loci is very  low.

No, I hold no great expectations for Zhiv either.  But, I do believe that the current estimates of time for many of our R-1b subclades is too short?  I am asking Why?  Note: it isn''t only my opinion here, its about half or more of the genetic "guru" community.  So, I am not Don Quixote out chasing windmills.  Reasonable minds have made the same observation.  I think the best chance to resolve the issue is "data mining", which I have some hands-on experience in.
« Last Edit: April 28, 2012, 11:24:36 AM by ironroad41 » Logged
Mike Walsh
Guru
*****
Offline Offline

Posts: 2963


WWW
« Reply #174 on: April 28, 2012, 11:32:42 AM »

As I remember markos work he also had an unexplained 5K to 6K years where he couldn't find enough data between L11 and subsequent mutations?
No, we went through that before. There is no inexplicable 5-6K years. It is just that at the L11 level of the tree there is no peer under L23 to do an interclade age with so he had to go up and over to an "uncle" outside of R-M269 for a valid haplogroup to compare with for interclade calculations.
« Last Edit: April 28, 2012, 11:33:18 AM by Mikewww » Logged

R1b-L21>L513(DF1)>L705.2
Pages: 1 ... 5 6 [7] 8 9 ... 14 Go Up Print 
« previous next »
Jump to:  


SEO light theme by © Mustang forums. Powered by SMF 1.1.13 | SMF © 2006-2011, Simple Machines LLC

Page created in 0.222 seconds with 17 queries.