Admixtools My qpAdm Results Please Share Yours!

[thunorr]

Since I use AT2 Qpadm, it’s in academicslop. I’ll translate it into human terms so people can better understand it.

Hunter-Gatherer and Farmer report:

42.1% ± 3% ANF
26.3% ± 2.77% EHG
10.9% ± 1.43% WHG
20.7% ± 3.59% CHG
P-Value: 0.891
Chi-Squared: 5.00

Steppe/ANF/WHG 3 way

48.8% ± 3.94% Western Steppe Herder
42.3% ± 3.65% ANF
8.86% ± 1.97% WHG
P-Value: 0.916
Chi-Squared: 4.60

The lefts and rights for my 0.916 P value Neolithic run:

Latvia_LN_CordedWare.AG,Greece_NeoNikomedeia_EN.SG,Luxembourg_Mesolithic.DG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Kostenki14_UP.SG,Russia_Sunghir3_UP.SG,Turkey_Central_Pinarbasi_Epipaleolithic.AG,Israel_Natufian.AG,Russia_Sidelkino_HG.SG,Georgia_Satsurblia_LateUP.SG,Switzerland_Bichon_Epipaleolithic.SG,Iran_Wezmeh_N.SG,Russia_MA1_UP.SG,ANE,China_AmurRiver_Paleolithic.AG

The rights and lefts for my 0.891 P-Value Mesolithic run:

Greece_NeaNikomedeia_EN.SG,Russia_Sidelkino_HG.SG,Luxembourg_Mesolithic.DG,Georgia_Satsurblia_LateUP.SG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Kostenki14_UP.SG,Russia_Sunghir3_UP.SG,Russia_MA1_UP.SG,Switzerland_Bichon_Epipaleolithic.SG,Israel_Natufian.AG,Turkey_Central_Pinarbasi_Epipaleolithic.AG,Turkey_Central_Boncuklu_PPN.AG,Italy_Epigravettian.AG.BY.AA,Iran_Wezmeh_N.SG,ANE,WSHG,Georgia_Dzudzuana_UP.SG

My index file:

https://www.mediafire.com/file/gp0ic0h427o2e1m/Thunorr_merged+4.ind/file

A FAIR WARNING, you might have to adjust models to suit your ethnic background. This model is very catered to my ethnic background, which is 98% northwest European.

 

[thunorr]

Since I use AT2 Qpadm, it’s in academicslop. I’ll translate it into human terms so people can better understand it.

Hunter-Gatherer and Farmer report:
View attachment 17992
42.1% ± 3% ANF
26.3% ± 2.77% EHG
10.9% ± 1.43% WHG
20.7% ± 3.59% CHG
P-Value: 0.891
Chi-Squared: 5.00

Steppe/ANF/WHG 3 way
View attachment 17991
48.8% ± 3.94% Western Steppe Herder
42.3% ± 3.65% ANF
8.86% ± 1.97% WHG
P-Value: 0.916
Chi-Squared: 4.60

The lefts and rights for my 0.916 P value Neolithic run:

Latvia_LN_CordedWare.AG,Greece_NeoNikomedeia_EN.SG,Luxembourg_Mesolithic.DG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Kostenki14_UP.SG,Russia_Sunghir3_UP.SG,Turkey_Central_Pinarbasi_Epipaleolithic.AG,Israel_Natufian.AG,Russia_Sidelkino_HG.SG,Georgia_Satsurblia_LateUP.SG,Switzerland_Bichon_Epipaleolithic.SG,Iran_Wezmeh_N.SG,Russia_MA1_UP.SG,ANE,China_AmurRiver_Paleolithic.AG

The rights and lefts for my 0.891 P-Value Mesolithic run:

Greece_NeaNikomedeia_EN.SG,Russia_Sidelkino_HG.SG,Luxembourg_Mesolithic.DG,Georgia_Satsurblia_LateUP.SG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Kostenki14_UP.SG,Russia_Sunghir3_UP.SG,Russia_MA1_UP.SG,Switzerland_Bichon_Epipaleolithic.SG,Israel_Natufian.AG,Turkey_Central_Pinarbasi_Epipaleolithic.AG,Turkey_Central_Boncuklu_PPN.AG,Italy_Epigravettian.AG.BY.AA,Iran_Wezmeh_N.SG,ANE,WSHG,Georgia_Dzudzuana_UP.SG

My index file:

https://www.mediafire.com/file/gp0ic0h427o2e1m/Thunorr_merged+4.ind/file

A FAIR WARNING, you might have to adjust models to suit your ethnic background. This model is very catered to my ethnic background, which is 98% northwest European.

Onto a deep ancestry model I made a while ago:

92.3% ± 2.52% UP West Eurasian
7.65% ± 2.52% UP East Eurasian
P-Value: 0.649
Chi-Squared: 5.99

An ANE model my friend made, I’m gonna consult him before releasing the model because I don’t want to unwillingly leak a model he worked hard on.

43.3% ± 7.35% Anatolian Neolithic Farmer
21.1% ± 1.51% Ancient North Eurasian
14.3% ± 2.29% Western Hunter-Gatherer
21.2% ± 9.04% Dzudzuana (not including the Dzudzuana ANF had)
P-Value: 0.428
Chi-Squared: 7.01

The ANE model is technically a fail, but right now I can’t be bothered to improve it.

Lefts and rights of the deep West/East Eurasian model:

Georgia_Kotias_UP.SG,China_AmurRiver_LPaleolithic.AG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Sunghir3_UP.SG,Russia_Kostenki14_UP.SG,ONG.SG,Laos_Hoabinhian.SG,Russia_Sunghir_UP.SG,Japan_Honshu_MidLateJomon.SG,Belgium_GoyetQ116_1_UP.AG,China_Tianyuan.AG.BY.AA

 

[Jalisciense]

Onto a deep ancestry model I made a while ago:

View attachment 17996

92.3% ± 2.52% UP West Eurasian
7.65% ± 2.52% UP East Eurasian
P-Value: 0.649
Chi-Squared: 5.99

An ANE model my friend made, I’m gonna consult him before releasing the model because I don’t want to unwillingly leak a model he worked hard on.

View attachment 17994

43.3% ± 7.35% Anatolian Neolithic Farmer
21.1% ± 1.51% Ancient North Eurasian
14.3% ± 2.29% Western Hunter-Gatherer
21.2% ± 9.04% Dzudzuana (not including the Dzudzuana ANF had)
P-Value: 0.428
Chi-Squared: 7.01

The ANE model is technically a fail, but right now I can’t be bothered to improve it.

Lefts and rights of the deep West/East Eurasian model:

Georgia_Kotias_UP.SG,China_AmurRiver_LPaleolithic.AG

Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Sunghir3_UP.SG,Russia_Kostenki14_UP.SG,ONG.SG,Laos_Hoabinhian.SG,Russia_Sunghir_UP.SG,Japan_Honshu_MidLateJomon.SG,Belgium_GoyetQ116_1_UP.AG,China_Tianyuan.AG.BY.AA

"21.2% ± 9.04% Dzudzuana (not including the Dzudzuana ANF had)"

The Georgia Kotias is not Dzudzuana but Caucasus Hunter-Gatherer.

 

[thunorr]

"21.2% ± 9.04% Dzudzuana (not including the Dzudzuana ANF had)"

The Georgia Kotias is not Dzudzuana but Caucasus Hunter-Gatherer.

Is it? I’m pretty sure it’s fully dzudzuana. Upper Paleolithic Caucasus. I could be wrong. I’ll model it later.

 

[qh777]

Upper paleolithic model for myself:

43.3% ± 7.35% Anatolian Neolithic Farmer
21.1% ± 1.51% Ancient North Eurasian
14.3% ± 2.29% Western Hunter-Gatherer
21.2% ± 9.04% Dzudzuana (not including the Dzudzuana ANF had)
P-Value: 0.428
Chi-Squared: 7.01

The ANE model is technically a fail, but right now I can’t be bothered to improve it.

unless you mean p value 0.0428, 0.428 is really good. Passing model seems to be p value 0.05 It's what I've read and seen commenters on this thread say.

 

[Jalisciense]

Is it? I’m pretty sure it’s fully dzudzuana. Upper Paleolithic Caucasus. I could be wrong. I’ll model it later.

Yes, we have already the Dzudzuana sample called Georgia_Dzudzuana_UP.

 

[thunorr]

Upper paleolithic model for myself:


View attachment 17997

unless you mean p value 0.0428, 0.428 is really good. Passing model seems to be p value 0.05 It's what I've read and seen commenters on this thread say.

Yes, 0.428
Post your results with the models I made.

Standard Error is also very important, I usually don’t pass a model unless it has a standard error of sub-5% or the admixture proportion is higher than the standard error.

 

[thunorr]

Upper paleolithic model for myself:

View attachment 17997

unless you mean p value 0.0428, 0.428 is really good. Passing model seems to be p value 0.05 It's what I've read and seen commenters on this thread say.

Also, my “ANE” group is AfontovaGora3 + Tajikistan_Mesolithic

 

[qh777]

Yes, 0.428
Post your results with the models I made.

Standard Error is also very important, I usually don’t pass a model unless it has a standard error of sub-5% or the admixture proportion is higher than the standard error.

You might be a bit too strict especially if the populations used in the model are reasonable. It depends on the assigned admixture. A standard error of 10% on an admixture of 40-50% seems alright to me. 5% standard error on 5% admixture isn't.

asked chatGPT about it:

 

[thunorr]

Yes, 0.428
Post your results with the models I made.

Standard Error is also very important, I usually don’t pass a model unless it has a standard error of sub-5% or the admixture proportion is higher than the standard error.

I was tired as I typed that, what I meant was:
“Or the standard error is higher than the admixture proportion”

 

[qh777]

Hey, Thunorr try some 2-way runs with Sweden IA 2 and Scotland MIA with the same samples I used like the Minoan, Mycenean, Chhim Phoenician, and Third Intermediate period Egyptian and see what your results are. And share them here of course!

Iron Age Greek + Sweden IA, Iron Age Greek + Scotland MIA

 

[qh777]

Steppe/ANF/WHG 3 way

48.8% ± 3.94% Western Steppe Herder
42.3% ± 3.65% ANF
8.86% ± 1.97% WHG
P-Value: 0.916
Chi-Squared: 4.60

Here's what I get with your 3 selected populations:

Same model with Iran_N:

The addition of Iran neolithic improves the fit for me. I'm curious to see if it improves yours too.

 

[thunorr]

Here's what I get with your 3 selected populations:

View attachment 18011

Same model with Iran_N:

View attachment 18012
The addition of Iran neolithic improves the fit for me. I'm curious to see if it improves yours too.

Try Israel_PPNB on the left too.

 

[thunorr]

Hey, Thunorr try some 2-way runs with Sweden IA 2 and Scotland MIA with the same samples I used like the Minoan, Mycenean, Chhim Phoenician, and Third Intermediate period Egyptian and see what your results are. And share them here of course!

Iron Age Greek + Sweden IA, Iron Age Greek + Scotland MIA

View attachment 18009
View attachment 18010

I’ll try it some time, my google colab is really bad and it crashes every time I try to import the .geno file, I’ll probably have to get a laptop and R studio or something.

 

[Jovialis]

It would cool to see what other model's you have ran.

Here is my ultimate source population admixture using an exact replica of the Raveane et al. 2022 methodology but with the AADR 62.0 dataset.

 

[thunorr]

View attachment 18013

Here is my ultimate source population admixture using an exact replica of the Raveane et al. 2022 methodology but with the AADR 62.0 dataset.

Try the model I made.

 

[thunorr]

Here's what I get with your 3 selected populations:

View attachment 18011

Same model with Iran_N:

View attachment 18012
The addition of Iran neolithic improves the fit for me. I'm curious to see if it improves yours too.

Oddly, the sample I used has WHG-like admixture. It’s just about 1%, but it’s enough to mess up the WHG percentage. My model is, however, quite literally so robust, it brute forces its way to a p value of 0.5 even when I use Yamnaya. In my old models, I had to use the highest steppe Corded Ware samples in order to get a p value that was passing.

48.6% ± 2.98% WSH
41.1% ± 2.37% ANF
10.2% ± 1.59% WHG
P-Value: 0.505
Chi-Squared: 9.29



I’ll model myself using the models you made earlier, the google colab runtime usage limit is about to kill me.


And, yes. My p-value does improve when I add Iran Neolithic, but only very slightly and it worsens SE for the other groups, plus I don’t carry any genetic ancestry from groups that would carry Zagrosian DNA.

44.7% ± 4.39% WSH
40.9% ± 2.35% ANF
11.0% ± 1.70% WHG
3.39% ± 2.79% Zagros
P-Value: 0.514
Chi-Squared: 8.20

 

[Jovialis]

Oddly, the sample I used has WHG-like admixture. It’s just about 1%, but it’s enough to mess up the WHG percentage. My model is, however, quite literally so robust, it brute forces its way to a p value of 0.5 even when I use Yamnaya. In my old models, I had to use the highest steppe Corded Ware samples in order to get a p value that was passing.


View attachment 18017
48.6% ± 2.98% WSH
41.1% ± 2.37% ANF
10.2% ± 1.59% WHG
P-Value: 0.505
Chi-Squared: 9.29



I’ll model myself using the models you made earlier, the google colab runtime usage limit is about to kill me.


And, yes. My p-value does improve when I add Iran Neolithic, but only very slightly and it worsens SE for the other groups, plus I don’t carry any genetic ancestry from groups that would carry Zagrosian DNA. View attachment 18016
44.7% ± 4.39% WSH
40.9% ± 2.35% ANF
11.0% ± 1.70% WHG
3.39% ± 2.79% Zagros
P-Value: 0.514
Chi-Squared: 8.20

Could you link the FAM/IND file, and paste the script?

 

[qh777]

Steppe/ANF/WHG 3 way

48.8% ± 3.94% Western Steppe Herder
42.3% ± 3.65% ANF
8.86% ± 1.97% WHG
P-Value: 0.916
Chi-Squared: 4.60

Here's what I get with your 3 selected populations:

View attachment 18011

Same model with Iran_N:

View attachment 18012
The addition of Iran neolithic improves the fit for me. I'm curious to see if it improves yours too.

Oddly, the sample I used has WHG-like admixture. It’s just about 1%, but it’s enough to mess up the WHG percentage. My model is, however, quite literally so robust, it brute forces its way to a p value of 0.5 even when I use Yamnaya. In my old models, I had to use the highest steppe Corded Ware samples in order to get a p value that was passing.


View attachment 18017
48.6% ± 2.98% WSH
41.1% ± 2.37% ANF
10.2% ± 1.59% WHG
P-Value: 0.505
Chi-Squared: 9.29



I’ll model myself using the models you made earlier, the google colab runtime usage limit is about to kill me.


And, yes. My p-value does improve when I add Iran Neolithic, but only very slightly and it worsens SE for the other groups, plus I don’t carry any genetic ancestry from groups that would carry Zagrosian DNA. View attachment 18016
44.7% ± 4.39% WSH
40.9% ± 2.35% ANF
11.0% ± 1.70% WHG
3.39% ± 2.79% Zagros
P-Value: 0.514
Chi-Squared: 8.20

I'd prefer using Anatolians from Turkey itself rather than Greece if the goal is to create a Steppe, ANF, WHG model.
Try Iran Ganj Dareh instead along with the Latvia LN corded ware to be a perfect comparison to my results.

 

[thunorr]

Could you link the FAM/IND file, and paste the script?

I’m using a version of qpAdm on google colab, but here is my index file:

https://www.mediafire.com/file/gp0ic0h427o2e1m/Thunorr_merged+4.ind/file

Lefts:
Russia_Samara_EBA_Yamnaya.AG,Greece_NeaNikomedeia_EN.SG,Luxembourg_Mesolithic.DG

Rights:
Mbuti.DG,Russia_UstIshim_IUP.DG,Russia_Kostenki14_UP.SG,Russia_Sunghir3_UP.SG,Turkey_Central_Pinarbasi_Epipaleolithic.AG,Israel_Natufian.AG,Russia_Sidelkino_HG.SG,Georgia_Satsurblia_LateUP.SG,Switzerland_Bichon_Epipaleolithic.SG,Iran_Wezmeh_N.SG,ANE,China_AmurRiver_Paleolithic.AG,Turkey_Marmara_Barcin_N.SG