we Kurds often get modelled as Druze + Lezgin . or similar

[Kurdquistador]

Druze + Lezgin

Lebanese + Lezgin

Iranian Jewish + Kumyk

etc.


sure the distances are not ideal most of the time . but i noticed that many kurds get modelled like this and many even with good fits !


anyway here is for example how i get modelled in Eurogenes EU test


55,08% Druze + 44,92% Lezgin @ 6,403


33% Mandean + 33% Druze + 33% Lezgin @ 5,267


--------------


Eurogenes K15 :

50% Lebanese_Druze + 50% Lezgin @ 6,954

50% Ossetian + 50% Jordanian @ 7,112

50% Lebanese_Christian + 50% Lezgin @ 7,232

50% Lebanese_Druze + 50% Adygei @ 7,321



-------------

Mdlp World Ancient Roots K18 :

50% Samaritians + 50% Lezgin @ 7,032

50% Druze + 50% Lezgin @ 7,387



----


Eurogenes K10

50% LB + 50% Kumyk @ 6,813

LB must mean Lebanese

50% Palestinian + 50% Chechen @ 7,544

Gedrosia DNA World 25 :

50% Syrian + 50% Lezgin @ 6,258
50% Kumyk + 50% Syrian @ 6,456
50% Chechen + 50% Syrian @ 6,601
50% Balkar + 50% Syrian @ 6,662
50% Lezgin + 50% Lebanese @ 6,713
50% Adygei + 50% Syrian @ 6,762

--

Harrappa World :

50% iraqi-arab + 50% kumyk @ 6,007
50% iraqi-arab + 50% north-ossetian @ 6,189
50% iraq-jew + 50% stalskoe @ 6,893
50% iraqi-arab + 50% lezgin @ 7,108
50% palestinian1 + 50% kumyk @ 7,150

--

GedrosiaDNA Eurasia K14 :

50% Kumyk + 50% Syrian @ 6,129

50% Adygei + 50% Syrian @ 6,162

---

Puntdnal K12 Near East :

50% Iraqi_Jew + 50% Lezgin @ 6,290
50% Syrian + 50% Lezgin @ 8,287


--------------------


all from Eurogenes k13

50% Ossetian + 50% Lebanese_Druze @ 5,850
50% Assyrian + 50% Kumyk @ 6,039
50% Kurdish_Jew + 50% Adygei @ 6,130
50% Kurdish_Jew + 50% Ossetian @ 6,328
50% Kurdish_Jew + 50% Balkar @ 6,394
50% Syrian + 50% Georgian @ 6,537

 

[FinalFlash]

Target: Kurdish Distance: 0.0161% / 0.01608895
63.7	Iranian_Mazandarani
36.3	Greek_Cappadocia

A much better fit^

Roughly 2/3 Northern Iranian+ 1/3 Anatolian Greek

 

[Kurdquistador]

of course but i want populations that are not too close to be used

by the way i didnt post G25 but gedmatch

 

[FinalFlash]

Iranian Mazandarani and especially Greek Cappadocia are hardly "close" to Kurds though.

 

[Kurdquistador]

well mazandaranis are "northwest iranic" . and anatolian greeks are from anatolia and not far from us geographically lol

also btw. we kurds can be modelled as 60% anatolian greek + 40% BMAC or something

 

[FinalFlash]

well mazandaranis are "northwest iranic" . and anatolian greeks are from anatolia and not far from us geographically lol

also btw. we kurds can be modelled as 60% anatolian greek + 40% BMAC or something

Yeah I meant they're not that close in a genetic sense.

 

[Kurdquistador]

Yeah I meant they're not that close in a genetic sense.

i consider up to distance 5 or 6 as close and up to 8-10 fairly close or "not far" personally

 

[FinalFlash]

i consider up to distance 5 or 6 as close and up to 8-10 fairly close or "not far" personally

That's a very interesting take to say the least, especially considering that differences already become quite noticeable even at distances of 2-3.

Personally, 5-6 is already fairly distant for me, never mind 8-10.

 

[FinalFlash]

Anything above 2.0% is bust. Even 1%-2% gedmatch/g25 can fail basic qpAdm.

For G25 and Gedmatch, anything above a 3 for a population AVERAGE is a bust for me. These Gedmatch models are laughably ridiculous.

 

[Kurdquistador]

Anything above 2.0% is bust. Even 1%-2% gedmatch/g25 can fail basic qpAdm.

what do you mean

 

[Kurdquistador]

For G25 and Gedmatch, anything above a 3 for a population AVERAGE is a bust for me. These Gedmatch models are laughably ridiculous.

then nobody is close to anything outside of their street . lel

 

[FinalFlash]

then nobody is close to anything outside of their street . lel

I meant in terms of modeling a population. A fit distance above 3 is inaccurate imo.

 

[FinalFlash]

The basic problems with PCAs are

a) overfitting, the algorithm will overfit by default in order to produce the 'closest' result. That's why I always limit pop references to 4 max and/or apply penalization=1 (as per the algorithm's inventor's instructions) to avoid overfitting as much as possible.

b) Poor/low count/single samples are treated as the same weight as high quality/multiple sample references. Ganj-Dareh has about 11 good references in Reich 1240K, Kotias is only 1. This is one of the reasons G25 CHG/Iran_N models cannot match qpAdm results. It's also one of the reasons why it's hard to produce working models with just Kotias in qpAdm (probably also the reason why academic studies usually opt for Iran_N).

So Turks (random example) celebrating that low East Asian % as proof of Turkic ancestry, they should investigate further before being too quick to count their spoils of war.


More on this problematic @ https://www.biorxiv.org/content/10.1101/2021.04.11.439381v4.full

How long would it take you to run a passing model?

 

[Kurdquistador]

I meant in terms of modeling a population. A fit distance above 3 is inaccurate imo.

i disagree

 

[FinalFlash]

@Onyx

I skimmed through the article you linked and I agree with it's main premise. It's too easy to manipulate results in these PCAs and oracles, and the Ashkenazi examples prove this. Personally, I've always been very skeptical about the "out of levant" theory for these people.