Mega QAPDM Models thread

Scottish + Sicilian

Attachment 142465


Scottish + North Italian

Attachment 142466



When modeling myself, the Scottish and North Italian run is the highest 2-way p-value so far. Even compared to Scottish and South French:

Attachment 142467

I've shared this model before, however, with this run I used a different set of references. P-value isn't as good as the previous model, but it's still comfortably above 0.05

Attachment 142473

References:

Attachment 142474
Attachment 142475

Quote:

---

Originally Posted by Gannicus

I've shared this model before, however, with this run I used a different set of references. P-value isn't as good as the previous model, but it's still comfortably above 0.05

Attachment 142473

References:

Attachment 142474
Attachment 142475

---

It might be a bit above 0,05 but it is still bad don't you think?

I might be to harsh but i would only take p values around 0,5 and above serious.

Quote:

---

Originally Posted by Croatian Legend

It might be a bit above 0,05 but it is still bad don't you think?

---

I was able to repeat this model earlier with a related Levant Neolithic population with different references:

Attachment 142476

This was the earlier run with a much higher p value:

Attachment 142477


Levant N populations have a significant Anatolian component. And Corded Ware is not 100% Western Steppe Herder.

Quote:

---

Originally Posted by Croatian Legend

I might be to harsh but i would only take p values around 0,5 and above serious.

---

That's not correct. https://www.nature.com/articles/s41586-025-09195-5

P-value for NUE001 that was modeled as ~80% Morocco MN and 20% Mesopotamia N yielded a p value of 0.12

"Ancestry of the Nuwayrat genome
We used the qpAdm42 framework to model the genetic ancestry components that best represent the Nuwayrat genome using a fully rotating model competition approach, in which a set of candidate populations are iteratively used as sources to construct one-source, two-source and three-source population ancestry models, whereas the remaining candidates are set as outgroup (right) populations43,44 (Supplementary Information section 4). We used a set of 13 populations from Neolithic and Chalcolithic West Asia, North Africa and the North Mediterranean region that predate the Nuwayrat individual as potential sources (Fig. 3c,d and Methods). No single-source model fitted the data (maximum P value observed = 2.39 × 10−6 for a model with Morocco_MN as a single source). Instead, a single two-source model (P = 0.12) met the significance criteria (P > 0.05), which consisted of a mixture of 77.6 ± 3.8% ancestry represented by genomes from the Middle Neolithic Moroccan site of Skhirat-Rouazi dated to 4780–4230 bce (Morocco_MN), and the remainder most closely related to genomes from 9000 to 8000 bce Neolithic Mesopotamia (22.4 ± 3.8%; Fig. 3a). In addition, two three-source models showed similar ancestry proportions but with a minor contribution of a third ancestry component represented by genomes from the Neolithic/Chalcolithic Levant (4.7 ± 8.2% at P = 0.11 and 1.1 ± 8.7% at P = 0.07, respectively; Supplementary Table 6)."

Quote:

---

Originally Posted by Gannicus

That's not correct. https://www.nature.com/articles/s41586-025-09195-5

P-value for NUE001 that was modeled as ~80% Morocco MN and 20% Mesopotamia N yielded a p value of 0.12

"Ancestry of the Nuwayrat genome
We used the qpAdm42 framework to model the genetic ancestry components that best represent the Nuwayrat genome using a fully rotating model competition approach, in which a set of candidate populations are iteratively used as sources to construct one-source, two-source and three-source population ancestry models, whereas the remaining candidates are set as outgroup (right) populations43,44 (Supplementary Information section 4). We used a set of 13 populations from Neolithic and Chalcolithic West Asia, North Africa and the North Mediterranean region that predate the Nuwayrat individual as potential sources (Fig. 3c,d and Methods). No single-source model fitted the data (maximum P value observed = 2.39 × 10−6 for a model with Morocco_MN as a single source). Instead, a single two-source model (P = 0.12) met the significance criteria (P > 0.05), which consisted of a mixture of 77.6 ± 3.8% ancestry represented by genomes from the Middle Neolithic Moroccan site of Skhirat-Rouazi dated to 4780–4230 bce (Morocco_MN), and the remainder most closely related to genomes from 9000 to 8000 bce Neolithic Mesopotamia (22.4 ± 3.8%; Fig. 3a). In addition, two three-source models showed similar ancestry proportions but with a minor contribution of a third ancestry component represented by genomes from the Neolithic/Chalcolithic Levant (4.7 ± 8.2% at P = 0.11 and 1.1 ± 8.7% at P = 0.07, respectively; Supplementary Table 6)."

---

It clear to me that everything above 0,05 can be looked at as valid. Still it is better to go for a higher p value if it is possible.

Quote:

---

Originally Posted by Gannicus

I was able to repeat this model earlier with a related Levant Neolithic population with different references:

Attachment 142476

This was the earlier run with a much higher p value:

Attachment 142477


Levant N populations have a significant Anatolian component. And Corded Ware is not 100% Western Steppe Herder.

---

Well they seem way better!

May i ask why you try to model yourself with natufian like components? Do you have ancestry from there? Just wondering, since i tried to create models with natufian in it for myself and it fails.

Quote:

---

Originally Posted by Croatian Legend

It clear to me that everything above 0,05 can be looked at as valid. Still it is better to go for a higher p value if it is possible.

---

Yes, I do aim for higher p values if I can. :)

There are other papers that publish models with p values of 0.06 too.

Lazaridis et al. 2022 paper on the southern arc: "The genetic history of the Southern Arc: A bridge between West Asia and Europe"