Some qpAdm models for Levänluhta

[Anglesqueville]

Just a quick thread to post some qpAdm models (Admixtools under Linux) for the two non-Germanic Finns from Levänluhta, Iron Age, DA234 and DA238. The motivation for this little research was, as is often the case, the right list. I know that some guys think they have found the philosopher's stone on this question. This is not my case, I am as hesitant and questioning as when I started, and this is even though as a mathematician I believe I have perfectly understood the theoretical background as Patterson posed it, and that I have also exchanged a lot on these questions with colleagues whose speciality is probabilistic theories. But from theoretical foundations to implementation there is often a long path, and for me, this path is often erratic. I therefore compared the models obtained with several right lists. The first (RL) was very poor in ancient Siberian populations. So I added some to it (see further).

RL=
Russia_Ust_Ishim.DG
Cameroon_SMA.DG
Italy_North_Villabruna_HG
Czech_Vestonice16
Belgium_UP_GoyetQ116_1
Russia_MA1_HG.SG
Iran_GanjDareh_N
Russia_Kostenki14.SG
Indian_GreatAndaman_100BP.SG
Israel_PPNB
Georgia_Kotias.SG
Turkey_N_I0707
Sidelkino.ial

RL1 = RL + Kolyma_meso.ial
RL2 = RL +  Kolyma_meso.ial + Amur_neo.ial
RL3 = RL + Kolyma_meso.ial + Amur_neo.ial + Yakutia_LN
RL4 = RL + Kolyma_meso.ial + Amur_neo.ial + Yakutia_LN + Russia_Krasnoyarsk_BA.SG

The models are under the form "Sweden_IA + 2d source", the 2d source being Minino_IA or BOO. All samples are post-imputation from Allentoft.
Whatever the right list the coefficients remain the same (taking the SE in account), and the SE remain around 0.03. The only question is that of the tail probabilities.

DA234 RL1 RL2 RL3 RL4
2d = Minino 0.74 0.68 0.74 0.51
2d = BOO 0.24 0.11 0.21 0.06

DA238 RL1 RL2 RL3 RL4
2d = Minino 0.45 0.45 0.34 0.38
2d = BOO 0.40 0.21 0.21 0.05

Your thoughts are welcome.

[Jaska]

I have not used this method and therefore I am not sure what numbers are what. But if I got it right, you gave the portions for BOO and Minino (as 1 - Sweden Iron Age)?

BOO suffers more from putting Siberian ancestries within the right populations? Which apparently means that there is not much left in BOO which is needed for the Levänluhta population, if the Siberian ancestry is removed. As Minino does not suffer equally, it would be more suitable as the true root population for Levänluhta? And this cannot be explained just from the fact that Minino has altogether less Siberian ancestry? But rather something like in Minino the Siberian ancestry has admixed with other ancestries and cannot be nullified so easily by adding "pure" Siberian ancestries within the right populations?

[Anglesqueville]

No, I gave the tail probs. The coefficients are never very different from these:

with right list: RL + Kolyma_meso.ial + Amur_neo.ial + Yakutia_LN + Russia_Krasnoyarsk_BA.SG

left pops:
Finland_IA.imputed_allentoft_234
Sweden_IA.imputed_allentoft
Russia_Minino_IronAge.ial


14.105        0.517555    0.479    0.521

left pops:
Finland_IA.imputed_allentoft_238
Sweden_IA.imputed_allentoft
Russia_Minino_IronAge.ial


15.957        0.384943    0.146    0.854


left pops:
Finland_IA.imputed_allentoft_234
Sweden_IA.imputed_allentoft
BOO.ial


24.115        0.063169    0.477    0.523

left pops:
Finland_IA.imputed_allentoft_238
Sweden_IA.imputed_allentoft
BOO.ial


24.510      0.0569241    0.159    0.841

Apologize if I was not explicit enough. My point was only the tail probs.

[Jaska]

Thank you for clarification!
OK, amazingly similar percentages, while so remarkable difference in tail probs. If I remember right, you still hesitate to make final conclusions even though the difference is clear?

Hmm... Concerning DA234, it looks that the tail prob gets higher when you add Yakutia (RL3), but drops again when you add Krasnoyarsk - for both Minino and BOO. But for DA238 it is the Krasnoyarsk which increases tail prob, and only concerning Minino. How do you interpret this? Could it hint to two different routes for the spread of the Siberian ancestry?

[Jalisciense]

Anyone who knows how to use qpAdm deserves my like and attention even if I have no idea what they are doing, writing or populations they are talking about.

[Anglesqueville]

If it was absolutely necessary to choose a single source for the Siberian component, my preference would without hesitation go to Minino_IA. But all studies have pointed to multiple waves of influx of this component, so these models are probably only crude simplifications for a complex phenomenon. In this regard, I am impatiently awaiting the publication of the study by Childebayeva, Haak et al. Why the introduction of kra001 in the right list ( RL4) collapses the tail probs of models with BOO as a Siberian source is not clear to me. While waiting for a convincing explanation, I will therefore avoid this choice.

[Anglesqueville]

Anyone who knows how to use qpAdm deserves my like and attention even if I have no idea what they are doing, writing or populations they are talking about.

You write "people who know how to use qpAdm". I'm not sure I can be counted among these people, but on the other hand I would advise you to be extremely wary of those who claim to be one (and there is no shortage of them). For my part I am simply in the twilight zone defined by a perfect understanding of the theoretical background (but between us it is not very difficult mathematics), and the annoying evidence that this understanding is of poor help in many cases. I add that I never use Admixtools2 (the R version).

[VladMC]

If it was absolutely necessary to choose a single source for the Siberian component, my preference would without hesitation go to Minino_IA. But all studies have pointed to multiple waves of influx of this component, so these models are probably only crude simplifications for a complex phenomenon. In this regard, I am impatiently awaiting the publication of the study by Childebayeva, Haak et al. Why the introduction of kra001 in the right list ( RL4) collapses the tail probs of models with BOO as a Siberian source is not clear to me. While waiting for a convincing explanation, I will therefore avoid this choice.

Because BOO is not a proper source? The real source is another line coming from the kra001 ?

[kolompar]

It means Levanluhta has affinity to kra001 that is not accounted for by Bolshoy (but it is by Minino).
I wonder why that is, because in terms of Y-DNA Bolshoy is consistent with descending from kra001, I would have thought it's the EHG/WSHG ancestry that's the difference. What happens if you add such populations to the right?

[Jalisciense]

Anyone who knows how to use qpAdm deserves my like and attention even if I have no idea what they are doing, writing or populations they are talking about.

You write "people who know how to use qpAdm". I'm not sure I can be counted among these people, but on the other hand I would advise you to be extremely wary of those who claim to be one (and there is no shortage of them). For my part I am simply in the twilight zone defined by a perfect understanding of the theoretical background (but between us it is not very difficult mathematics), and the annoying evidence that this understanding is of poor help in many cases. I add that I never use Admixtools2 (the R version).

Thanks for the advice, but my comment was half joking and half serious xd

[VladMC]

It means Levanluhta has affinity to kra001 that is not accounted for by Bolshoy (but it is by Minino).
I wonder why that is, because in terms of Y-DNA Bolshoy is consistent with descending from kra001, I would have thought it's the EHG/WSHG ancestry that's the difference. What happens if you add such populations to the right?

This fully corresponds to the archaeological data from the European part of Russia. Since 2000 BC, archaeology has exclusively shown movement from the Volga to the north and there is absolutely no evidence of movement from the north to the Volga. At the same time, according to archaeological data, the Ural-speaking population lived up to the southern Urals.

[Jaska]

VladMC:

At the same time, according to archaeological data, the Ural-speaking population lived up to the southern Urals.

Actually archaeological data cannot tell that. Archaeological data is just a match or not a match for the linguistic results. Perhaps you meant that.

[Norfern-Ostrobothnian]

I don't think the Indo-European contacts with PU can be used to locate the location of PU either. In fact I think constraining PU and para-Uralic languages in a single small location is very dubious since the lifestyles in this area were quite mobile in the Bronze Age.
I'd say PU and other Uralic related languages inhabited quite a wide range even before the extant branches of Uralic split apart.

[Jaska]

I don't think the Indo-European contacts with PU can be used to locate the location of PU either. In fact I think constraining PU and para-Uralic languages in a single small location is very dubious since the lifestyles in this area were quite mobile in the Bronze Age.
I'd say PU and other Uralic related languages inhabited quite a wide range even before the extant branches of Uralic split apart.

You seem to have severely misunderstood something. 

1. Of course contacts can be used to locate Proto-Uralic, because languages cannot be in contact unless they are spoken in adjacent regions. 

2. Every language and proto-language are born in a narrow homeland. They cannot be born in a wide region, because in that case the result would not be a uniform language at all. 

3. How narrow the homeland actually was, depends on the livelihood and the number of speakers. Farmer societies lived densely, while pastoralists moved long distances during a year, as did hunter-gatherers. However, we know from the Taiga zone in Siberia and North America that typical hunter-gatherer language community was/is no greater than few hundreds or thousands of speakers. 

4. In the Lapp villages in Northern Fennoscandia, which were around 10 000 km2 (100 x 100 km or 50 x 200 km), there lived ca. 200 hunter-gatherers in each Lapp village during the 16th-17th centuries. Still there had been born an own dialect in every Lapp village. That corresponds well to a homeland of ancient proto-language, spoken by hunter-gatherers in the northern Taiga zone.

[Norfern-Ostrobothnian]

1. Indo-European languages were spoken on such a wide range at the time the loanwords were introduced to Uralic that placing them to specifically the Urals isn't necessary
Especially considering the only realistic options are extinct branches of Indo-European, Indo-Iranian or Tocharian languages.
2. Languages can also emerge within a larger area and then become the only extant member of the previous language family, which is what happened with Yeniseian languages. Same could have happened with Proto-Uralic, forming out of a larger group of which only it remains.
3. Consider the modern language distributions of Siberia. Evenk is spoken from Krasnoyarsk oblast to the Amur basin. And this has persisted for hundreds of years. And again I'm not saying the language didn't disintegrate multiple times before becoming the immediate finaly Proto-Uralic form, but I doubt it was restricted to a single region with no Para-Uralic relatives around for thousands of years either
4. Calling Sami hunter-gatherers is kind of dubious especially in the 16th-17th century. And consider that the region used to stretch all the way to Dvina Karelia and persisted for two thousand years or so. Even so, Lapland is quite large in on itself, with moder Sapmi being aroun 400 000 km2