Allentoft and Iron Age Sweden

[Codaman]

^^ I haven't been interested in NEO538 at all so far, so I don't have anything to say about it. As for using VolgaOka_IA as a source for the Saamis (including the two from Levänluhta) this is a perfectly justified choice if one follows Lang's theories (which is my case, at least on this point). This choice requires to be completed by a bonus from Krasnoyarsk_BA, and gives very satisfactory adjustments with qpAdm.

Per the theories of Lang NEO538 represents the northeastern group of what he calls Tapiola Ware, which is in linguistuics putatively associated with certain similarities of Saami and Merya to the opposition to the more western Baltic Finnic.
It would be difficult to justify using something far more remote than it in my opinion.

[Anglesqueville]

^^Codaman:
Well, I'll respond properly (at least, I hope) to your post #11, precisely to the G25 model that you propose for DA234, which attributes to DA234 a significant affinity with Falköping_NEO221 (27%). I propose this model to qpAdm (still via Admixtools latest version, qpfstats, allSNPs=YES), and here is the result:
left pops:
Finland_IA.imputed_allentoft_da234
Russia_IronAge.ial
Sweden_N_Falköping.imputed_Allentoft_NEO221
Estonia_BA_ial

best coefficients: 0.492 0.127 0.382
totmean: 0.492 0.127 0.382
boot mean: 0.491 0.128 0.382
std. errors: 0.120 2.773 2.847


fixed pat wt dof chisq tail prob
000 0 12 11.464 0.489586 0.492 0.127 0.382
001 1 13 16.994 0.199571 0.550 0.450 0.000
010 1 13 12.334 0.500455 0.487 0.000 0.513
100 1 13 65.445 5.47297e-09 0.000 -12.120 13.120 infeasible
011 2 14 76.942 1.04015e-10 1.000 0.000 0.000
101 2 14 118.751 1.10487e-18 0.000 1.000 0.000
110 2 14 115.660 4.43534e-18 0.000 0.000 1.000
best pat: 000 0.489586 - -
best pat: 010 0.500455 chi(nested): 0.870 p-value for nested model: 0.351058
best pat: 011 1.04015e-10 chi(nested): 64.608 p-value for nested model: 9.13983e-16

2 remarks:
1) the standard errors are astronomical. In particular the standard error attached to the coefficient of NEO221 is 2.776 for a coefficient of 0.127. This alone in my opinion would be enough to condemn this model. But there is worse:
2) the p-value of the WITHOUT NEO221 model (nested model 010) largely passes the fateful bar: 0.35.

As a check, if I propose the model WITHOUT NEO221, I obtain a very satisfactory model, both for its correct standard errors and for its adjustment:

best coefficients: 0.487 0.513
totmean: 0.487 0.513
boot mean: 0.488 0.512
std. errors: 0.052 0.052

fixed pat wt dof chisq tail prob
00 0 13 12.317 0.501807 0.487 0.513
01 1 14 76.647 1.17869e-10 1.000 0.000
10 1 14 116.034 3.75014e-18 0.000 1.000
best pat: 00 0.501807 - -
best pat: 01 1.17869e-10 chi(nested): 64.330 p-value for nested model: 1.05252e-15

For my part, I have long abandoned G25 for serious questions. This case is a perfect demonstration of the reasons which led me to this abandonment

[Codaman]

I appreciate the effort Angles, but I proposed earlier NEO225 as the Falköping source rather than NEO221 as I noticed there was a a massive difference in their coverage(0.06 vs 4.50). Perhaps I should have made a note about that.
In any case, I do believe this genome NEO538 is a skeleton key of sorts to uncovering who was was living in Finland in the late bronze and early Iron Age through comparison with Levänluhta. People akin to Bronze Age Estonians seem very likely perhaps in combination with people originating from Scandinavia.

[Anglesqueville]

This post is in fact off-topic since this thread had no other motivation than to propose the model present in my first post. But since the discussions made NEO538 intervene, which is new to me, and it turns out that NEO538 has just given me a new model for Levänluhta that is much better than anything I had seen before, I am taking advantage of this thread to publish it. Warnings:
1) I am completely ignorant of the archaeological background of NEO538, so I don't know whether this model can have any reality beyond the numerical one.
2) Motala intervenes (very very modestly). Motala's data is imputed, but by me (Beagle5). I will try later with the Motalas imputed by Allentoft. I doubt there will be big differences, but you never know. So here is this model:

left pops:
Finland_IA.imputed_allentoft (Levänluhta)
VolgaOka_IA
Russia_IronAge.ial (NEO538)
Motala_Hg_imp

best coefficients: 0.481 0.479 0.040
totmean: 0.481 0.479 0.040
boot mean: 0.480 0.480 0.040
std. errors: 0.078 0.066 0.031

fixed pat wt dof chisq tail prob
000 0 12 7.833 0.79803 0.481 0.479 0.040
001 1 13 10.535 0.649748 0.529 0.471 0.000
010 1 13 69.780 8.8129e-10 1.024 0.000 -0.024 infeasible
100 1 13 38.075 0.000280481 0.000 0.866 0.134
011 2 14 70.491 1.57267e-09 1.000 0.000 0.000
101 2 14 53.597 1.50455e-06 0.000 1.000 0.000
110 2 14 1009.891 0 0.000 0.000 1.000
best pat: 000 0.79803 - -
best pat: 100 0.000280481 chi(nested): 30.242 p-value for nested model: 3.81322e-08
best pat: 101 1.50455e-06 chi(nested): 15.521 p-value for nested model: 8.15812e-05

[Codaman]

This post is in fact off-topic since this thread had no other motivation than to propose the model present in my first post. But since the discussions made NEO538 intervene, which is new to me, and it turns out that NEO538 has just given me a new model for Levänluhta that is much better than anything I had seen before, I am taking advantage of this thread to publish it. Warnings:
1) I am completely ignorant of the archaeological background of NEO538, so I don't know whether this model can have any reality beyond the numerical one.
2) Motala intervenes (very very modestly). Motala's data is imputed, but by me (Beagle5). I will try later with the Motalas imputed by Allentoft. I doubt there will be big differences, but you never know. So here is this model:

Regarding the context. The linguistic hypothesis for Levänluhta is that their ancestors came from the vicinity of Vologda/southeastern Ladoga between 1300 and 200 BC with the upper end being the start of distinctively Volga-affiliated cultural influences in non-coastal Finland and the latter corresponding closely with the age of NEO538. There appears no compelling argument against using it.
If there is additional involvement of Volga-Oka it may have something to do with the middle passage that Lang postulates which may have brought a different type of ancestry from NEO538.


Also by the way these two Levänluhta genomes you use do not necessarily represent the same population. DA238 had a terrestial diet while DA234 had a marine diet.

the 87Sr/86Sr data suggest, that individual JK1963/DA238 was probably a local relying on terrestrial foodstuffs, while individuals JK1968/DA234 and JK2067/DA237 might well have been locals incorporating a component of Baltic sea resources in their diet.

I am not quite yet willing to concede that DA234 had no Scandinavian admixture as the technical issues involved leave the evidence uncertain. Saami people even in Norway who had migrated to their present lands by or during the times of Levänluhta carry the same I1 subclades as Finns do.

[Norfern-Ostrobothnian]

This post is in fact off-topic since this thread had no other motivation than to propose the model present in my first post. But since the discussions made NEO538 intervene, which is new to me, and it turns out that NEO538 has just given me a new model for Levänluhta that is much better than anything I had seen before, I am taking advantage of this thread to publish it. Warnings:
1) I am completely ignorant of the archaeological background of NEO538, so I don't know whether this model can have any reality beyond the numerical one.
2) Motala intervenes (very very modestly). Motala's data is imputed, but by me (Beagle5). I will try later with the Motalas imputed by Allentoft. I doubt there will be big differences, but you never know. So here is this model:

left pops:
Finland_IA.imputed_allentoft (Levänluhta)
VolgaOka_IA
Russia_IronAge.ial (NEO538)
Motala_Hg_imp

best coefficients:    0.481    0.479    0.040
totmean:      0.481    0.479    0.040
boot mean:    0.480    0.480    0.040
      std. errors:    0.078    0.066    0.031

fixed pat  wt  dof    chisq      tail prob
          000  0    12    7.833        0.79803    0.481    0.479    0.040
          001  1    13    10.535        0.649748    0.529    0.471    0.000
          010  1    13    69.780      8.8129e-10    1.024    0.000    -0.024  infeasible
          100  1    13    38.075    0.000280481    0.000    0.866    0.134
          011  2    14    70.491    1.57267e-09    1.000    0.000    0.000
          101  2    14    53.597    1.50455e-06    0.000    1.000    0.000
          110  2    14  1009.891              0    0.000    0.000    1.000
best pat:          000          0.79803              -  -
best pat:          100      0.000280481  chi(nested):    30.242 p-value for nested model:    3.81322e-08
best pat:          101      1.50455e-06  chi(nested):    15.521 p-value for nested model:    8.15812e-05

Use EHG instead of Motala because the archaelogical cultures of Finland were more related to Russian than to Swedish Mesolithic.

[Codaman]

Use EHG instead of Motala because the archaelogical cultures of Finland were more related to Russian than to Swedish Mesolithic.

Based on what I have heard the nearly contemporary individuals from the Luistari site 200 km south of Levänluhta had SHG or Baltic HG admixture rather than EHG. Their autosomal relation with modern day Finns and Saami is obscure still but some of the things said about them are suggestive of them having Siberian admixture somewhat exceeding modern Finnish variation.
This HG could mean they were simply rich in BA Estonian ancestry or it could mean something else, we might find out later this year. With G25 there is little need for any additional EHG source as NEO538 has that covered.

[Anglesqueville]

A two-source model VolgaOka + NEO538 gives very clean results for each of the two Levänluhta individuals, with reduced standard errors and more than correct tail probs. Such a model may not be an exhaustive description of reality, but it cannot be very far from it. As for an HG bonus, I completely ignored it. Whether it is SHG or EHG, it is extremely weak at best anyway. As for a possible Germanic influx (or proto- if you prefer), the addition of Falköping or Sweden_IA to the sources leads in one case to an infeasible model, in the other to a "rotten" model (huge standard errors greater than the coefficients). It is to such "rotten" models that the addition to the sources of Estonia_BA leads. If we follow Aikio in his analysis of Baltic > Saami lexical borrowings (all probably according to him mediated by Finnish) this is not surprising. Now obviously qpAdm is just qpAdm, and a search for shared IBDs might perhaps shed some light on this. I will leave you to meditate on the 2-source models, and on the interesting inversion of balance that they present.

left pops:
Finland_IA.imputed_allentoft_DA234
VolgaOka_IA
Russia_IronAge.ial

best coefficients: 0.755 0.245
totmean: 0.755 0.245
boot mean: 0.754 0.246
std. errors: 0.078 0.078


fixed pat wt dof chisq tail prob
00 0 13 11.062 0.605632 0.755 0.245
01 1 14 22.348 0.0717404 1.000 0.000
10 1 14 77.367 8.68449e-11 0.000 1.000
best pat: 00 0.605632 - -
best pat: 01 0.0717404 chi(nested): 11.286 p-value for nested model: 0.000780882


left pops:
Finland_IA.imputed_allentoft_DA238
VolgaOka_IA
Russia_IronAge.ial


est coefficients: 0.305 0.695
totmean: 0.305 0.695
boot mean: 0.301 0.699
std. errors: 0.088 0.088


fixed pat wt dof chisq tail prob
00 0 13 9.390 0.74287 0.305 0.695
01 1 14 93.929 6.83512e-14 1.000 0.000
10 1 14 20.481 0.115679 0.000 1.000
best pat: 00 0.74287 - -
best pat: 10 0.115679 chi(nested): 11.091 p-value for nested model: 0.000867415

[Queequeg]

Many thanks for your efforts, Angles.

[Anglesqueville]

Anecdotally, the model with Finland_IA restricted to DA234 is a little better ( if one assumes that an increase of 0.02 on the tail_prob has some consistency, which is disputable):

left pops:
Sweden_IA.imputed_allentoft
Sweden_N_Falköping.imputed_Allentoft
Finland_IA.imputed_allentoft_DA234


best coefficients: 0.896 0.104
totmean: 0.896 0.104
boot mean: 0.896 0.104
std. errors: 0.043 0.043

fixed pat wt dof chisq tail prob
00 0 13 5.981 0.946827 0.896 0.104
01 1 14 12.105 0.597834 1.000 0.000
10 1 14 136.257 3.92742e-22 0.000 1.000
best pat: 00 0.946827 - -
best pat: 01 0.597834 chi(nested): 6.124 p-value for nested model: 0.0133354

Anyway, everyone on here is aware that the matter of such models is not human groups nor archaeological cultures but alleles lists. Be it DA234 or DA234+DA238, this "source" points to a certain population which for evident reasons is not Levänluhta_IA itself, is perhaps even not genealogically directly related to Levänluhta, but shares with it a large number of alleles. I could write the same for "Sweden_IA", which is on this model 3 genomes (Rise174 + 2 from Margaryan). I plan to apply this model to the Iron Age people from Mattila's study, we'll see.
note: I hope that when reading this model everyone thinks to have a look at the p-value of the nested 01: 0.01, largely below the fatal threshold.

[Anglesqueville]

An error had crept into the .ind file of my working file. The Sweden_IA group did indeed contain VK523, which is Norwegian. On the other hand, VK534 was missing. Here is the model with the right individuals. Basically nothing different. The tail_prob is less impressive (why? I have no idea). On the other hand, this one is even clearer (even no need to look at the p-value of the alternative model, its tail_prob is ridiculous).

left pops:
Sweden_IA.imputed_allentoft
Sweden_N_Falköping.imputed_Allentoft
Finland_IA.imputed_allentoft_DA234

best coefficients:    0.858    0.142
totmean:      0.858    0.142
boot mean:    0.858    0.142
      std. errors:    0.042    0.042

fixed pat  wt  dof    chisq      tail prob
          00  0    13    11.585        0.561904    0.858    0.142
          01  1    14    23.425      0.0536996    1.000    0.000
          10  1    14  134.515    8.69947e-22    0.000    1.000
best pat:          00        0.561904              -  -
best pat:          01        0.0536996  chi(nested):    11.839 p-value for nested model:    0.000579894


I'm taking this opportunity to make a quick clarification, since it seems that this post aroused the attention (and aggression) of someone on a well-known blog. Everyone must have noticed that I did not draw any linguistic conclusions from these models. More precisely, I resisted this temptation. If the Scandinavian individuals involved had been a few centuries older, it would undoubtedly have been different. What can we speculate from these models? I'll let you think about it. For my part, my job here was simply to publish models that were on my machine, period.

[jdean]

I'm taking this opportunity to make a quick clarification, since it seems that this post aroused the attention (and aggression) of someone on a well-known blog. Everyone must have noticed that I did not draw any linguistic conclusions from these models. More precisely, I resisted this temptation. If the Scandinavian individuals involved had been a few centuries older, it would undoubtedly have been different. What can we speculate from these models? I'll let you think about it. For my part, my job here was simply to publish models that were on my machine, period.

Take no notice, he's a bitter individual.

[Jaska]

I'm taking this opportunity to make a quick clarification, since it seems that this post aroused the attention (and aggression) of someone on a well-known blog. Everyone must have noticed that I did not draw any linguistic conclusions from these models. More precisely, I resisted this temptation. If the Scandinavian individuals involved had been a few centuries older, it would undoubtedly have been different. What can we speculate from these models? I'll let you think about it. For my part, my job here was simply to publish models that were on my machine, period.

People often tend to get "snow-blind" with the linguistic labels: after they see white, all they see is white. 

The more we get ancient DNA-samples, the more we can find traces of contacts between different regions; and the same goes with the archaeological data. This means that the probability that a random genetic or cultural influence would be associated with the prevailing language lineage in a certain region diminishes continuously.

There could have been many language shifts (even back and forth) in every region before the prevailing language finally became dominant there. Nobody can see these language shifts or expansions in the DNA or in the archaeological data, even though many still believe that they can.

[Anglesqueville]

Many thanks for your efforts, Angles.

Queequeg, I've read your comments on Eurogenes. Note that I never claimed that the Falköping_LN group carried a "Uralic" affinity. Apart from the fact that historically I cannot imagine the possibility, I have never published any model targeting Falköping. Everything I say concerns exclusively the handful of individuals reporting to Sweden_IA (see the correction made to their list). Regarding these, if only the PCA were at stake, the conclusion would be uncertain, to say the least. Unlike others, I am never satisfied with a single reading of the first two components of a PCA. It is not by chance that I decided on qpAdm modeling. That qpAdm highlights an oriental influx for this Iron Age micro-group by taking Falköping as the main source seems to me difficult to contest, at least numerically. As for the interpretation of this numerical fact, everyone noted that I offered none.

[Rodoorn]

If the forefathers of modern I1 diverged apart from OST003 perhaps during the Ertebølle Culture at said time they would probably be primarily WHG. Perhaps when farming was introduced into Northern Europe the forefathers of Modern I1 moved to northeastern Scandinavia to completely avoid the Funnelbeaker Culture as opposed to the forefathers of Mr. OST003 who maintained their hunter-gathering lifestyle and genetics despite being surrounded by the farmers in Northern Germany for around 800 years?

I guess a few questions I still have are when and where did the forefathers of Modern I1 encounter the Corded Ware Culture? By the time they show up in Falköping during the Late Neolithic they have primarily Corded Ware autosomal dna correct? Did they have early contact with the Corded Ware somewhere in the Eastern Baltic? At said point how did they maintain their I1 patrilineal lines despite absorbing primarily Corded Ware autosomal dna?

According to Karsten Wentink in his dissertation of 2006, the spread of TRB on the North German Plain- originated about the Ostorf area- was in fact 'powered by' Ertebølle HG, that had partly took a neolithic package.

According to Egfjørd in his paper of 2021 SGC spread towards the North German Plain and Jutland about 2850 BC, from the middle Elbe area (he mentions Halle in Central-east Germany).