World Families Forums - R-L21 and Subclades Variance results

Welcome, Guest. Please login or register.
September 21, 2014, 10:00:15 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)
| | |-+  R-L21 and Subclades Variance results
« previous next »
Pages: [1] Go Down Print
Author Topic: R-L21 and Subclades Variance results  (Read 1212 times)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« on: September 30, 2012, 07:46:04 PM »

With the Busby Study I calculated 15 STR (14 actually) variances for the various locations, I decided to run the 67 marker L21 and predicted haplotypes from the L21 spreadsheet dated 09292012 and the main subclades with descending subclades as I though might be interesting. Using my modified MikeW Age Estimator with the new KenN Gen111T and the latest MarkoH mutation rates Ken used and 389b rate edit.


Edit: Will repost as variance was incorrectly calculated for this run.

Ok, here is the revision due to formula error in using a prior beta spreadsheet. Re-ran all the results and split the list into Subclades. Of course the variance rankings changed.

https://docs.google.com/file/d/0By9Y3jb2fORNaUV5dW1jRDRqU0E/edit

Sorry, I dont have any research assistants to review things. :)

MJost
« Last Edit: October 02, 2012, 11:38:14 AM by Mark Jost » 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)
Mike Walsh
Guru
*****
Offline Offline

Posts: 2964


WWW
« Reply #1 on: October 01, 2012, 09:56:28 AM »

With the Busby Study I calculated 15 STR (14 actually) variances for the various locations, I decided to run the 67 marker L21 and predicted haplotypes from the L21 spreadsheet dated 09292012 and the main subclades with descending subclades as I though might be interesting. Using my modified MikeW Age Estimator with the new KenN Gen111T and the latest MarkoH mutation rates Ken used and 389b rate edit.

The top three or so Populations may need to be ignored due to only several haplotypes as it appears the low number affects the results but three of more work. England East maybe the most variant. Sweden is significant. England has several other population area are in the top levels, along with DF13** England.

IF most of the English surnames are not Continental types, then this is very interesting results.

https://docs.google.com/open?id=0By9Y3jb2fORNWlZfVTJwd0NIaHM

MJost

Thanks, Mark. I noticed (because I'm it) that for the L705.2 group you show a count of only 2. How did you select the two? I think there are seven or eight 67 STR L705.2 confirmed folks.

How did you use the mutation rates in the calculations? It appears you are sorting this by STR variance. STR variance calculations do not use mutation rates in and of themselves. Did you adjust the variance by the mutation rate of the STR?
Logged

R1b-L21>L513(DF1)>S6365>L705.2(&CTS11744,CTS6621)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #2 on: October 01, 2012, 01:01:27 PM »

Deleted. Found spread sheet formula error. Will rework.
« Last Edit: October 01, 2012, 02:42:15 PM by Mark Jost » 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)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #3 on: October 01, 2012, 01:35:10 PM »


How did you use the mutation rates in the calculations? It appears you are sorting this by STR variance. STR variance calculations do not use mutation rates in and of themselves. Did you adjust the variance by the mutation rate of the STR?

No, I did not adjust any thing. I only mentioned that the Gen111T had Marko's mutation rates used to calculate the Ages which is variance divided by the sum of mutation rate for the specific marker panel used.

Edit: You must have noticed variance issue which I just rechecked and found a formula error in the spreadsheet. I will work all the variance.
MJost
« Last Edit: October 01, 2012, 02:43:45 PM by Mark Jost » 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)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #4 on: October 01, 2012, 10:30:54 PM »

With the Busby Study I calculated 15 STR (14 actually) variances for the various locations, I decided to run the 67 marker L21 and predicted haplotypes from the L21 spreadsheet dated 09292012 and the main subclades with descending subclades as I though might be interesting. Using my modified MikeW Age Estimator with the new KenN Gen111T and the latest MarkoH mutation rates Ken used and 389b rate edit.

The top three or so Populations may need to be ignored due to only several haplotypes as it appears the low number affects the results but three of more work. England East maybe the most variant. Sweden is significant. England has several other population area are in the top levels, along with DF13** England.

IF most of the English surnames are not Continental types, then this is very interesting results.

https://docs.google.com/open?id=0By9Y3jb2fORNWlZfVTJwd0NIaHM

MJost

Thanks, Mark. I noticed (because I'm it) that for the L705.2 group you show a count of only 2. How did you select the two? I think there are seven or eight 67 STR L705.2 confirmed folks.


Mike, I reran your subclade and here are the results

Count   Founder's Age   Generations   StdDevInGen   YBP   + -YBP   Max   VAR   SD
N=4   L21 DF13/L513/L706.2/L705.2 England   41.2   18.7   1,235.8   562.1   1,797.9   4.8333   2.198
N=15   L21 DF13/L513/L706.2/L705.2 All   37.7   17.9   1,129.9   537.5   1,667.3   4.4190   2.102
N=3   L21 DF13/L513/L706.2/L705.2 Wales   31.2   16.3   937.5   489.6   1,427.1   3.6667   1.915


I am reworking the L21. Next time I will make sure I have the correct version be before I start my projects... Dag nabbit.

MJost
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)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #5 on: October 02, 2012, 11:38:48 AM »


Ok, here is the revision due to formula error in using a prior beta spreadsheet. Re-ran all the results and split the list into Subclades. Of course the variance rankings changed.

https://docs.google.com/file/d/0By9Y3jb2fORNaUV5dW1jRDRqU0E/edit
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)
alan trowel hands.
Guru
*****
Offline Offline

Posts: 2012


« Reply #6 on: October 02, 2012, 01:55:24 PM »

Your results puzzle me.  I would like to hear comments from Mike etc on why they seem to differ from theirs. 
Logged
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #7 on: October 02, 2012, 02:33:13 PM »

Your results puzzle me.  I would like to hear comments from Mike etc on why they seem to differ from theirs. 

You mean Busby's data?

MJost
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)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #8 on: October 02, 2012, 06:08:09 PM »

I added some DF13+ varieties to the existing chart.

https://docs.google.com/file/d/0By9Y3jb2fORNaUV5dW1jRDRqU0E/edit

MJost

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)
Heber
Old Hand
****
Offline Offline

Posts: 448


« Reply #9 on: October 02, 2012, 10:31:52 PM »

I added some DF13+ varieties to the existing chart.

https://docs.google.com/file/d/0By9Y3jb2fORNaUV5dW1jRDRqU0E/edit

MJost



Very interesting analysis Mark. I noted that Iberia has the highest variance for L21 in Europe, which would support the model I have in mind. As a general rule I don't have a lot of confidence in STR age or variance estimates as they vary widely and have covered the whole spectrum of the peopling of Europe. However the analysis tools appear to be getting better.
Logged

Heber


 
R1b1a2a1a1b4  L459+ L21+ DF21+ DF13+ U198- U106- P66- P314.2- M37- M222- L96- L513- L48- L44- L4- L226- L2- L196- L195- L193- L192.1- L176.2- L165- L159.2- L148- L144- L130- L1-
Paternal L21* DF21


Maternal H1C1



Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #10 on: October 03, 2012, 09:34:41 AM »



Very interesting analysis Mark. I noted that Iberia has the highest variance for L21 in Europe, which would support the model I have in mind. As a general rule I don't have a lot of confidence in STR age or variance estimates as they vary widely and have covered the whole spectrum of the peopling of Europe. However the analysis tools appear to be getting better.

I was very surprised with this result as well. Two things.

1) The 22 surnames are mostly or all Spanish AND MikeW has 20 of them categorized as 'Iberian Peninsula':

Sánchez, Rodriguez, Sánchez, Guerra, Rodriguez, Calzada, Romero, Amuchástegui, Sampedro, DeJesus, Lenares, Montoya, De la Puerta, Leal, Roma, Escalante, Gonsalus, Montanez, Olazabal, Cantu, Fernándes, Palmas

2) The 22 HTs are across six subclades or varieties. Now if this result had only single digit counts, I would be slighly tapped on the shoulder as bias due to size. I tried to test variance in these 22 guys using variance combinations and found one possible issue with 251256 Sánchez's  STRs 413's with 14,14 (are correct) producing a variance of 29.9. I pulled it and the variance dropped significantly but still a large variance of 23.9. Also the generation age drop to 203.7  down from 254.6 placing it in a more reasonable number as compared to the next few entries.

FYI: 251256  Francisco SÁNCHEZ, born 1798, Montilla (Córdoba)

Also there are several Isles location in both L21 and DF13 entries on the list that have top variances as well. It maybe the result of Atlantic coast hugging influences.

MJost

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)
Richard Rocca
Old Hand
****
Offline Offline

Posts: 523


« Reply #11 on: October 03, 2012, 09:42:01 AM »



Very interesting analysis Mark. I noted that Iberia has the highest variance for L21 in Europe, which would support the model I have in mind. As a general rule I don't have a lot of confidence in STR age or variance estimates as they vary widely and have covered the whole spectrum of the peopling of Europe. However the analysis tools appear to be getting better.

I was very surprised with this result as well. Two things.

1) The 22 surnames are mostly or all Spanish AND MikeW has 20 of them categorized as 'Iberian Peninsula':

Sánchez, Rodriguez, Sánchez, Guerra, Rodriguez, Calzada, Romero, Amuchástegui, Sampedro, DeJesus, Lenares, Montoya, De la Puerta, Leal, Roma, Escalante, Gonsalus, Montanez, Olazabal, Cantu, Fernándes, Palmas

2) The 22 HTs are across six subclades or varieties. Now if this result had only single digit counts, I would be slighly tapped on the shoulder as bias due to size. I tried to test variance in these 22 guys using variance combinations and found one possible issue with 251256 Sánchez's  STRs 413's with 14,14 (are correct) producing a variance of 29.9. I pulled it and the variance dropped significantly but still a large variance of 23.9. Also the generation age drop to 203.7  down from 254.6 placing it in a more reasonable number as compared to the next few entries.

FYI: 251256  Francisco SÁNCHEZ, born 1798, Montilla (Córdoba)

Also there are several Isles location in both L21 and DF13 entries on the list that have top variances as well. It maybe the result of Atlantic coast hugging influences.

MJost



In this case, wouldn't an "Iberia-All" bucket be more useful?
Logged

Paternal: R1b-U152+L2*
Maternal: H
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #12 on: October 03, 2012, 10:48:18 AM »


In this case, wouldn't an "Iberia-All" bucket be more useful?

Good point. So here is the analysis. The largest variance is 'Spain only - EW Iberian Peninsula' then Spain and Portugal - EW Iberian Peninsula .


MikeW Spreadsheet 09292012
n=5764 -  67 Markers dataset

Note: Iberia = Spain and Portugal

Spain only - EW Iberian Peninsula
n=20
31.69210526
without 251256 Sánchez
25.1871345

EW Iberian Peninsula
n=27
27.74643875
without 251256 Sánchez
22.63230769

EW Aquitaine & Pyrenees
EW Iberian Peninsula
n=29
26.75862069
without 251256 Sánchez
21.99470899

EW Aquitaine & Pyrenees
n=2
13.5
without 251256 Sánchez NA

Portugal Only - EW Iberian Peninsula
n=6
10.2
without 251256 Sánchez NA



ALL Country's included:

EW Iberian Peninsula
27.74643875
n=37
without 251256 Sánchez
22.63230769

EW Alpine & Cisalpine
EW Iberian Peninsula
26.41856061
n=33
without 251256 Sánchez
22.09576613

EW Aquitaine & Pyrenees
EW Iberian Peninsula
24.30930931
n=37
without 251256 Sánchez
20.45873016

EW Alpine & Cisalpine
EW Aquitaine & Pyrenees
EW Iberian Peninsula
23.75415282
n=43
without 251256 Sánchez
20.37514518

EW Alpine & Cisalpine
EW Aquitaine & Pyrenees
17.52222222
n=10
without 251256 Sánchez NA

MJost
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)
Mike Walsh
Guru
*****
Offline Offline

Posts: 2964


WWW
« Reply #13 on: October 03, 2012, 11:03:06 AM »

  STRs 413's with 14,14 (are correct) producing a variance of 29.9.

My recommendation is to throw out all multi-copy STRs for this kind of variance calculation. Vineviz recommended this to me some time ago.  Eliminating multi-copy STRs helps reduce the jumpiness (multi-step increments) that causes erratic results.

I'd also suggest that if removing or adding just one or two haplotypes makes a difference in the results you don't have enough data.
Logged

R1b-L21>L513(DF1)>S6365>L705.2(&CTS11744,CTS6621)
alan trowel hands.
Guru
*****
Offline Offline

Posts: 2012


« Reply #14 on: October 03, 2012, 12:31:43 PM »

I am puzzled why the results give a totally different picture to other like Mike's in terms of the geography of variance.  Maybe Mike could weigh in and comment?  I am in no position to say what is right an wrong as I am not mathematical
Logged
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #15 on: October 03, 2012, 06:44:38 PM »

  STRs 413's with 14,14 (are correct) producing a variance of 29.9.

My recommendation is to throw out all multi-copy STRs for this kind of variance calculation. Vineviz recommended this to me some time ago.  Eliminating multi-copy STRs helps reduce the jumpiness (multi-step increments) that causes erratic results.

I'd also suggest that if removing or adding just one or two haplotypes makes a difference in the results you don't have enough data.

Ken has not used 11 STRs: 385,389i,459,464,CDY (I changed the mutation rate that Ken used for389ii to 389ii-i rate and confirmed MarkoH's rate for 389b.) I already emailed and asked Ken about not using the remaining multicopy's and he didnt respond directly to the this part of my inquiry but wrote 'These files are periodically updated as new information is obtained.'   

Pulling out the remaining multi-copy markers: YCA II, 395S1 and 413 in Age calculation could be easily done leaving 52 STRs.

Test                 Multi-Copy Markers
====                 ================== 
12 Marker            385a, 385b
25 Marker Upgrade    459a, 459b and 464a, 464b, 464c, 464d
37 Marker Upgrade    YCA II a, YCA II b and CDY a, CDY b
67 Marker Upgrade    395S1a, 395S1b and 413a, 413b

MJost
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)
Mike Walsh
Guru
*****
Offline Offline

Posts: 2964


WWW
« Reply #16 on: October 03, 2012, 08:39:42 PM »

 STRs 413's with 14,14 (are correct) producing a variance of 29.9.

My recommendation is to throw out all multi-copy STRs for this kind of variance calculation. Vineviz recommended this to me some time ago.  Eliminating multi-copy STRs helps reduce the jumpiness (multi-step increments) that causes erratic results.

I'd also suggest that if removing or adding just one or two haplotypes makes a difference in the results you don't have enough data.

Ken has not used 11 STRs: 385,389i,459,464,CDY (I changed the mutation rate that Ken used for389ii to 389ii-i rate and confirmed MarkoH's rate for 389b.) I already emailed and asked Ken about not using the remaining multicopy's and he didnt respond directly to the this part of my inquiry but wrote 'These files are periodically updated as new information is obtained.'   

Pulling out the remaining multi-copy markers: YCA II, 395S1 and 413 in Age calculation could be easily done leaving 52 STRs.

Test                 Multi-Copy Markers
====                 ==================  
12 Marker            385a, 385b
25 Marker Upgrade    459a, 459b and 464a, 464b, 464c, 464d
37 Marker Upgrade    YCA II a, YCA II b and CDY a, CDY b
67 Marker Upgrade    395S1a, 395S1b and 413a, 413b

MJost

I won't defend anyone's TMRCA calculator, although I'd pick Ken's versus others just because Ken is really, really intelligent on this stuff.

However, the tool needs to be "fit for purpose".  I don't think he built it to calculate deep, deep ancestral relationships like 5,000 or 10,000 years ago.  Bennett Greenspan has said FTDNA's tip TMRCA calculator is only good for about 2000 years... or rather he wouldn't use it beyond that..  

The real reason that I avoid building my own TMRCA calculator and try to avoid the discussion is because it all depends on the mutation rates and no one has conclusively proved what rates are the right rates.

Ken told me for my purposes, relative age, I should not use diversity. It does not have a straight line correlation with time. He said use variance... I asked if I should normalize across STRs and both he and Vince V told me use just straight sum of the variance.

Most people that have problems with all of this, not necessarily Ken and Vince V, cite STR saturation. This is much of the Busby study's concern about about a Neolithic origin in Europe for R1b-L11/S127. Ken acknowledges it exists and I think Vince would too.  Based on this current I've tried to evaluate STR selection.

I think it is pretty clear that all multicopy markers and possible nulls should be thrown out. I don't want to rehash all this again and again. Most of it is over on the STR Wars thread.

Quote from: Ken Nordtvedt 12/29/10
I, personally, have my doubts that GD in the genealogical time frame, or maybe any time frame, is so sharp a tool to warrant including these multi-copy markers. There are too many other issues such as statistical nature of mutation, accounting for back mutations, uncertainty about the actual mutational rules --- that render GD a rough tool. I do, however, use these markers (459 and 464) in simply recognizing distinct clades and clusters.

Quote from: Vince Vizachero
The problem is that the underestimation does not scale linearly. I'm not sure we ever fully discussed in on the list or not, but I recall that we did about a year ago.  What's happening is that the variance for the markers with the highest mutation rates is not scaling with time as quickly as the slower markers.  In short, the fastest markers underestimate TMRCA (in my opinion) progressively more at longer TMRCAs. If, in addition to omitting the multi-copy markers, you omit the five or ten fastest mutating ones
I think you'll find that your TMRCA estimates using interclade variance are much more reasonable. The problem isn't interclade estimation:  the problem is that highly mutative STR \s appear to not accumulate variance linearly when the time period is long.

Quote from: John Chandler 07/03/10
Every marker experiences the same process of dispersal, but at its own characteristic rate.  In short, each marker can be used as an independent estimator for TMRCA with its own uncertainty.  The reason one might wish to exclude multi-copy markers is that, except in close relationships, they may add more uncertainty than they take away, since the various copies are not individually identified by the testing. That doesn't mean they bias the estimates if they are included.

I personally tested some of these things by trying to run various sets of haplotypes and across various haplogroups where the haplogroup relationships are known. For example, we know L2 is younger than U152, L21 younger than P312, etc.  I saw consistent results when using large numbers of STRs, like 30 or more, and without using multi-copy.  An actuarist on the M222 forum says you really should have 50 STRs.  

In terms of sample size, my experience is that usually I needed at least 20 haplotypes per group, but real stability sets in somewhere above 30 and by 50 the die seems to always have been cast. Adding or subtracting haplotypes doesn't change the outcome. Of course, we always want a representative sample and Richard R will be the first to say our FTNDA project data is biased. He's right, but the academic data is biased too. There are gaping holes in the data, period. I personally think the long haplotype, deep SNP tested project data is very valuable, just not an end all, and obviously with large gaps in the data.  A couple of quick examples are Romania (i.e. real Romanians) and Denmark which is one I don't understand (our lack of.)

Variance across geographies is more risky than across haplogroups. We know the haplogroups are related to each other more closely into the comparative group. We can not say that about geographic groups.
« Last Edit: October 04, 2012, 12:23:23 AM by Mikewww » Logged

R1b-L21>L513(DF1)>S6365>L705.2(&CTS11744,CTS6621)
Mark Jost
Old Hand
****
Offline Offline

Posts: 707


« Reply #17 on: October 03, 2012, 11:25:28 PM »

I posted results of using 17 vs 11 original Multi-copy markers that Ken N used in his Gen111T.

The overall sum of mutation rates and Variance decreases. The original concern I had with the 111 marker age was less than the 67 marker result is now almost identical, Using 111 still improves the confidence level of course.

https://docs.google.com/file/d/0By9Y3jb2fORNTDU0WjlXbGRZbVk/edit

I will update my beta version with the removal of the remaining multi-copy markers.

MJost.
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)
Mike Walsh
Guru
*****
Offline Offline

Posts: 2964


WWW
« Reply #18 on: October 04, 2012, 09:00:43 AM »

I am puzzled why the results give a totally different picture to other like Mike's in terms of the geography of variance.  Maybe Mike could weigh in and comment?  I am in no position to say what is right an wrong as I am not mathematical

We should remember that the primary rationale behind the concept of L21 not being from Spain was that the P312 in Spain appeared to not be L21. There is not much L21 from Spain, the sample size is building though. Variances with small sample sized and limited STRs should be treated cautiously and that's what we've had for Spain.

I just ran a variance with the 36 linear marker set and, yes, similar to Mark, I did get Spain with a higher variance than either France or the Isles. That was only a sample of 20 67 STR haplotypes so I'm a little nervous of concluding much. I'll reopen the Spanish R-L21 thread because this is definitely worth investigation. There needs to be more testing from this area. There are some strange haplotypes from Spain, or at least of guys with Spanish surnames and origin countries of Spain.

Alan, did you know there is P314.2 look-alike a guy from Spain? We need to get him tested.  There was an L513 look alike guy but he came up L513-. At the time wondered if this wasn't just an early branch just DF13+ but pre-L513. I don't know. The guy needs to test for DF13.

I'll recheck the MDKA origins and probably go to a thread just to look at these. As I've noted, its not a long list. I think it makes a difference if they are all from the Pyrenees region versus the rest of Spain.... or really just Spanish surnames without genealogical knowledge of a connect to Spain.
« Last Edit: October 04, 2012, 09:03:43 AM by Mikewww » Logged

R1b-L21>L513(DF1)>S6365>L705.2(&CTS11744,CTS6621)
Pages: [1] 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 0.097 seconds with 18 queries.