Mega QAPDM Models thread

Printable View

Show 40 post(s) from this thread on one page

10-07-2025, 12:57 AM
Peterski

Any idea what I'm doing wrong that I'm getting results like this with every model?:

(p values are good but Feasibility is always FALSE and there are crazy percentages)

https://i.postimg.cc/ZKS9zrKL/example.png
10-07-2025, 02:33 AM
Opie

Quote:

Originally Posted by Peterski

Any idea what I'm doing wrong that I'm getting results like this with every model?:

(p values are good but Feasibility is always FALSE and there are crazy percentages)

https://i.postimg.cc/ZKS9zrKL/example.png

I assume you chose significantly overlapping groups.
10-07-2025, 01:20 PM
Opie

Quote:

Originally Posted by Peterski

(p values are good but Feasibility is always FALSE and there are crazy percentages)

https://i.postimg.cc/ZKS9zrKL/example.png

Choose only one Scythian sample in your references.
10-07-2025, 02:06 PM
drb234

Quote:

Originally Posted by Gannicus

I did another Steppe/ANF/WHG model of myself. The goal here was to maximize p-value and tighten up the standard errors as much as possible.

Attachment 143054

References:

Attachment 143055
Attachment 143056
Attachment 143057

Low Steppe for a NW Euro
10-07-2025, 10:55 PM
Gannicus

Quote:

Originally Posted by drb234

Low Steppe for a NW Euro

If you want you can give me a different list of reference populations and see what that does.
10-07-2025, 11:03 PM
drb234

Quote:

Originally Posted by Gannicus

If you want you can give me a different list of reference populations and see what that does.

Try this
Lefts: "Russia_Samara_EBA_Yamnaya.AG,Greece_NeaNikomedeia _EN.SG,Luxembourg_Mesolithic.DG"

Rights: "Mbuti.DG,Ju_hoan_North.DG,Russia_MA1_UP.SG,Russia _Steppe_Eneolithic.AG,Iran_Luristan_PPN.SG,Georgia _Satsurblia_LateUP.SG,Russia_DevilsCave_N.SG,Italy _Epigravettian.SG,Turkey_Central_Boncuklu_PPN.AG"

I dont know if it's a good model but it gave me alright results.
https://i.imgur.com/Le8kBnU.png
10-08-2025, 02:43 AM
Gannicus

2 Attachment(s)

Quote:

Originally Posted by drb234

Try this
Lefts: "Russia_Samara_EBA_Yamnaya.AG,Greece_NeaNikomedeia _EN.SG,Luxembourg_Mesolithic.DG"

Rights: "Mbuti.DG,Ju_hoan_North.DG,Russia_MA1_UP.SG,Russia _Steppe_Eneolithic.AG,Iran_Luristan_PPN.SG,Georgia _Satsurblia_LateUP.SG,Russia_DevilsCave_N.SG,Italy _Epigravettian.SG,Turkey_Central_Boncuklu_PPN.AG"

I dont know if it's a good model but it gave me alright results.
https://i.imgur.com/Le8kBnU.png

With your listed references:

Attachment 143685

With Barcin:

Attachment 143686
10-09-2025, 02:02 AM
drb234

neolithic model on the french
https://i.imgur.com/B12dOef.png
10-09-2025, 05:03 AM
Gannicus

1 Attachment(s)

Steppe/ANF/WHG admix model for this Gaul from central France:

Attachment 143696
10-10-2025, 09:50 PM
Gannicus

2 Attachment(s)

A Sinstashta model of myself with drb's reference populations:

Attachment 143726

Attachment 143727
10-13-2025, 10:59 PM
Xavier727

1 Attachment(s)

England_LIA / Denmark_EarlyViking

Attachment 143776
10-31-2025, 11:53 PM
Xavier727

6 Attachment(s)

England_MIA / Norway_IA

Attachment 144061

Attachment 144062
Attachment 144063
Attachment 144064

Playing around with the reference pops and managed to get the P value up slightly

Attachment 144078

Attachment 144077
11-01-2025, 01:28 AM
Gannicus

4 Attachment(s)

Quote:

Originally Posted by Xavier727

England_MIA / Norway_IA

Attachment 144061

Attachment 144062
Attachment 144063
Attachment 144064

Here's what I get:

Attachment 144065

Attachment 144066
Attachment 144067
Attachment 144068

I haven't tried it with your references.
11-01-2025, 11:44 AM
Xavier727

Quote:

Originally Posted by Gannicus

Here's what I get:

Attachment 144065

Attachment 144066
Attachment 144067
Attachment 144068

I haven't tried it with your references.

Yeah, you should try it out when you get a chance, I’d be interested to see what you get.
11-01-2025, 12:26 PM
Fistora

this is what I get when using your NW populations as a source combined with Russia Sunghir, which seems to me to be one of the best proxies for Slavs

https://i.postimg.cc/wvnfCNY0/image.png
11-02-2025, 02:20 AM
Gannicus

2 Attachment(s)

Quote:

Originally Posted by Xavier727

Yeah, you should try it out when you get a chance, I’d be interested to see what you get.

With your references:

Attachment 144086

With the additional references, it does not improve the p value for me.

Attachment 144087
11-02-2025, 11:30 AM
Xavier727

Quote:

Originally Posted by Gannicus

With your references:

Attachment 144086

With the additional references, it does not improve the p value for me.

Attachment 144087

Interesting, that’s quite a solid amount of Germanic you’ve got there, how much do you normally get with these IA models.
11-02-2025, 12:04 PM
AnthrogenicaRefugee

A guy from Russia ran my Ancestry kit.

Models that passed
https://i.postimg.cc/5HLD7gVg/IMG-20...103802-054.jpg

https://i.postimg.cc/NymZJDcC/IMG-20...103807-159.jpg

https://i.postimg.cc/CZk9P7g2/IMG-20...103816-345.jpg

Models with Thracian and Illyrian sources failed, only the Glinoe Scythian, which he said is actually Dacian worked.
11-02-2025, 07:03 PM
Gannicus

12 Attachment(s)

Quote:

Originally Posted by Xavier727

Interesting, that’s quite a solid amount of Germanic you’ve got there, how much do you normally get with these IA models.

It varies based on the samples used:

Attachment 144105

Attachment 144106

Attachment 144107

Attachment 144108

Not really sure that in most cases qpAdm can reliably distinguish between Celtic and Germanic populations.

I can model myself with just Celtic sources also:

Attachment 144109

Attachment 144110

Attachment 144111

Attachment 144112

Attachment 144113

And then Modeling with just Germanic sources:

Attachment 144114

Attachment 144115

Attachment 144116
11-02-2025, 08:47 PM
Xavier727

Quote:

Originally Posted by Gannicus

It varies based on the samples used:

Attachment 144105

Attachment 144106

Attachment 144107

Attachment 144108

Not really sure that in most cases qpAdm can reliably distinguish between Celtic and Germanic populations.

I can model myself with just Celtic sources also:

Attachment 144109

Attachment 144110

Attachment 144111

Attachment 144112

Attachment 144113

And then Modeling with just Germanic sources:

Attachment 144114

Attachment 144115

Attachment 144116

Yeah, I've noticed that as well, which is interesting given how people always say how much better qpAdm is to G25. Also, it seems that your choice of right populations can have quite a big impact on your results, which is annoying given that there doesn't seem to be any clear formula on what makes for an optimal reference set.
11-04-2025, 06:05 PM
Gannicus

1 Attachment(s)

I'm going to be experimenting with this combination more and see if I can get these standard errors down:

Attachment 144177
11-05-2025, 09:12 PM
Fistora

https://i.postimg.cc/43Ky3cXK/kddjsfjsfjsjf.png
11-07-2025, 08:38 PM
Gannicus

3 Attachment(s)

About the best I can do with this combination

Attachment 144257

Tried VK579 instead

Attachment 144258

Both were ran with the Gernarchivist references I think Opie linked in an earlier post.

Attachment 144259
11-14-2025, 02:18 AM
Artzenlohe
Quote:
Originally Posted by Lard

PCA vs. qpAdm: More Similar Than People Think

Fundamentally, autosomal PCA and qpAdm modeling operate on the same basic idea:
They both assess how populations are related based on genome-wide allele frequencies — just using different statistical approaches.

PCA uses principal components to reduce genetic variance into 2D/3D plots, revealing clustering patterns.
qpAdm estimates ancestry proportions by modeling a test population as a mix of chosen “left” sources, with “right” reference groups to anchor the model.

But in both cases, your results depend almost entirely on your inputs.
You can cherry-pick populations in either method — and get misleading results that still “look” scientific.

Why qpAdm Isn't Immune to Narrative Bias

There’s a tendency to treat qpAdm as more rigorous than PCA, and while it can be more formalized, it’s just as vulnerable to biased modeling choices.

That’s why so many published studies claim Ashkenazi Jews are “50%+ Levantine” — even though PCA consistently places them overlapping with Southern Italians.

Here’s why:
Most of these studies use deeply flawed European proxies — Northern Italians, French, Sardinians, or even Tuscans — while completely ignoring Southern Italians or Sicilians. This artificially inflates the Levantine component, because these proxies are genetically too distant and don't match PCA placement.

The qpAdm model still returns a “feasible” result — because qpAdm doesn't know your sources are bad.
It just fits whatever you give it.

In other words: Bad source selection in qpAdm doesn’t get corrected by the math — it gets rewarded. The tool will give you a p-value and ancestry breakdown that looks clean, but is fundamentally built on cherry-picked assumptions.

PCA-Informed qpAdm is the Solution

That’s why my models start with PCA — which clearly shows Ashkenazi Jews clustering directly on top of Southern Italians — and then use qpAdm to test models that reflect that structure.

Examples from my thread:
Stable qpAdm Models of Ashkenazi Jewish Ancestry – Ignoring PCA No More

Models such as:

Italian_South.HO + Lebanese_Muslim.HO
Italian_South.HO + Israel_Ashkelon_LBA.AG
Italian_South.HO + Israel_C_o2.AG

All produce stable 2-way models with:

~75–78% Southern Italian–like ancestry
~22–25% Levantine
Strong p-values and standard errors
Full consistency with PCA and historical records

This matches not just the genetic structure, but also the well-documented Roman and Byzantine-era Jewish presence in Southern Italy.

Visual Results

https://i.ibb.co/M0ZfVqC/1.png
Modern Levantine Proxy model

https://i.ibb.co/S4FnRm4s/2.png
Bronze Age Coastal Levantine (Philistine-admixed)

https://i.ibb.co/G43Kn5wk/3.png
Inland Bronze Age Levantine (likely Israelite/Canaanite)

Don’t Let Narrative Modeling Skew the Data

You can go check nearly any study by Behar, Hammer, or Ostrer — they almost all use Northern Italians or French as the European side of the model.
Even YouTubers like Andrei DNA — who is also an Apricity forum member — repeat the same setup and get the same distorted results: Ashkenazi Jews as a “Middle Eastern” population.

But if Ashkenazi Jews sit right on top of Southern Italians in PCA — and they do — then they’re no more “Middle Eastern” than a Calabrian is.

That’s why PCA-aligned qpAdm is essential — and narrative-driven modeling needs to be called out.
https://www.youtube.com/watch?v=5gxr79HfpYQ
You need southern Italians, Levantines, and Northern Europeans to model Ashkenazi Jews.
11-23-2025, 12:17 AM
Gannicus

3 Attachment(s)

Using AndreiDNA's right and left populations, I modeled my Basal Eurasian admixture.

https://www.youtube.com/watch?v=t59yXBo1MoI

Attachment 144721
Attachment 144724

Mbuti is the proxy for Basal Eurasian.
11-23-2025, 01:59 AM
userNa

Quote:

Originally Posted by Gannicus

Mbuti is the proxy for Basal Eurasian.

Bad proxy in my opinion.

Target: Israel_Natufian
Distance: 27.4820% / 0.27482024
97.4 Russia_Kostenki14
2.6 Mbuti

Target: Georgia_Kotias.SG
Distance: 28.6263% / 0.28626346
100.0 Russia_Kostenki14

Target: Iran_GanjDareh_N
Distance: 27.1517% / 0.27151689
99.8 Russia_Kostenki14
0.2 Mbuti
11-23-2025, 03:09 AM
Gannicus

Quote:

Originally Posted by userNa

Bad proxy in my opinion.

Target: Israel_Natufian
Distance: 27.4820% / 0.27482024
97.4 Russia_Kostenki14
2.6 Mbuti

Target: Georgia_Kotias.SG
Distance: 28.6263% / 0.28626346
100.0 Russia_Kostenki14

Target: Iran_GanjDareh_N
Distance: 27.1517% / 0.27151689
99.8 Russia_Kostenki14
0.2 Mbuti

You can't use G25 like that. It has to do with the outgroups that Andrei selected in his video. He made that video on it 4 months ago. Perhaps he has changed his mind on the method. It'd be nice if he could comment on it.
11-23-2025, 03:16 AM
userNa

Quote:

Originally Posted by Gannicus

-

I've seen his videos.

In my opinion what shows up as Mbuti in his model is just an out-of-africa wave that postdates the first Eurasian arrivals and not actual African admixture.
11-23-2025, 03:27 AM
Gannicus

Quote:

Originally Posted by userNa

I've seen his videos.

In my opinion what shows up as Mbuti in his model is just an out-of-africa wave that postdates the first Eurasian arrivals and not actual African admixture.

Right, it's not African at all. It's his attempt to measure Basal Eurasian that's using Mbuti as a proxy. Using his model with his source populations and reference populations it's speculated that I have 24% basal eurasian admixture. I thought it was interesting. So, I decided to share it.
11-23-2025, 03:39 AM
Gannicus

2 Attachment(s)

Anyways back to the more typical modeling. Using Xaviers references a few posts ago I modeled myself with 2 Iron age sources, Norway IA and Czechia IA Hallstatt.

Attachment 144728

Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0326% / 0.03258002
66.6 Czech_IA_Hallstatt
33.4 Norway_IA.SG

Attachment 144729

Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0326% / 0.03258002
66.6 Czech_IA_Hallstatt
33.4 Norway_IA.SG
11-23-2025, 03:50 AM
userNa

Quote:

Originally Posted by Gannicus

Right, it's not African at all. It's his attempt to measure Basal Eurasian that's using Mbuti as a proxy. Using his model with his source populations and reference populations it's speculated that I have 24% basal eurasian admixture. I thought it was interesting. So, I decided to share it.

Andrei claims (unless I haven't understood his videos) that Basal Eurasian is a politically correct term for African admixture in Eurasians. Anyways, thanks for posting.
11-23-2025, 04:01 AM
Gannicus

Quote:

Originally Posted by userNa

Andrei claims (unless I haven't understood his videos) that Basal Eurasian is a politically correct term for African admixture in Eurasians. Anyways, thanks for posting.

I think he has since changed his mind.
11-23-2025, 05:20 AM
Gannicus

4 Attachment(s)

Quote:

Originally Posted by Gannicus

Anyways back to the more typical modeling. Using Xaviers references a few posts ago I modeled myself with 2 Iron age sources, Norway IA and Czechia IA Hallstatt.

Attachment 144728

Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0326% / 0.03258002
66.6 Czech_IA_Hallstatt
33.4 Norway_IA.SG

Attachment 144729

Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0326% / 0.03258002
66.6 Czech_IA_Hallstatt
33.4 Norway_IA.SG

Tried it with Austria IA La Tene with different references this time.

Attachment 144730

Attachment 144731
Attachment 144732
Attachment 144733

And this is what it looks like in G25:

Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0309% / 0.03092630
55.0 Norway_IA.SG
45.0 Austria_IA_LaTene
11-23-2025, 07:17 AM
userNa

That's me:
p value: 0.0954|Feasibility: TRUE
Greece_Peloponnese_N.AG_I5427.AG 56.5 +/- 9.25%
Iran_GanjDareh_N.AG_I1954.AG 22.96 +/- 4.60 %
Russia_Samara_HG.AG_I0124.AG 8.43 +/- 4.67 %
Russia_Ekven_OldBeringSea.AG_I7333.AG 2.48 +/- 2.35%
Jordan_PPNC.AG_I1699.AG 9.62 +/- 8.62%
11-23-2025, 08:37 PM
Gannicus

Quote:

Originally Posted by userNa

That's me:
p value: 0.0954|Feasibility: TRUE
Greece_Peloponnese_N.AG_I5427.AG 56.5 +/- 9.25%
Iran_GanjDareh_N.AG_I1954.AG 22.96 +/- 4.60 %
Russia_Samara_HG.AG_I0124.AG 8.43 +/- 4.67 %
Russia_Ekven_OldBeringSea.AG_I7333.AG 2.48 +/- 2.35%
Jordan_PPNC.AG_I1699.AG 9.62 +/- 8.62%

Have you tried IA, BA, and Medieval yet?
11-23-2025, 08:46 PM
Xavier727

4 Attachment(s)

Austria_IA_LaTene.AG + Denmark_IA.SG

Attachment 144743

Attachment 144744
Attachment 144745
Attachment 144746

slightly different reference set
11-23-2025, 09:04 PM
userNa

Quote:

Originally Posted by Gannicus

Have you tried IA, BA, and Medieval yet?

I haven't managed to obtain a good model for BA and IA so far (standard error is too high).
I have a decent one for medieval though:

p value: 0.7524|Feasibility: TRUE|
Ukraine_Medieval_Rus.SG_VK541.SG 16.56 +/- 7.25%
Turkey_Marmara_Iznik_Basilica_Roman_Byzantine_A.AG_I8366.AG 79.06 +/- 11.67%
Egyptian.HO_Egypt7.HO 4.38 +/- 7.75%
11-23-2025, 09:44 PM
Gannicus

2 Attachment(s)

Quote:

Originally Posted by Xavier727

Austria_IA_LaTene.AG + Denmark_IA.SG

Attachment 144743

Attachment 144744
Attachment 144745
Attachment 144746

slightly different reference set

I used your previous two reference lists:

Attachment 144747

Attachment 144748
11-23-2025, 09:50 PM
Gannicus

1 Attachment(s)

Quote:

Originally Posted by userNa

I haven't managed to obtain a good model for BA and IA so far (standard error is too high).
I have a decent one for medieval though:

p value: 0.7524|Feasibility: TRUE|
Ukraine_Medieval_Rus.SG_VK541.SG 16.56 +/- 7.25%
Turkey_Marmara_Iznik_Basilica_Roman_Byzantine_A.AG_I8366.AG 79.06 +/- 11.67%
Egyptian.HO_Egypt7.HO 4.38 +/- 7.75%

might want to use ancient egyptians jk2134 and jk2911. Modern Egyptians have more SSA than their ancient ancestors.

Pooling the samples together might help too. Like this:

Attachment 144749
11-23-2025, 10:38 PM
Gannicus

3 Attachment(s)

More with England MIA LIA

Attachment 144750

Attachment 144751

Attachment 144752

Show 40 post(s) from this thread on one page