MDLP K13 Ultimate Gedmatch calculator's weirdness

[Dora]

Hi all,

I have already asked a similar question sometime before about another test, but this test is even stranger for me. Does anybody can explain what MDLP K13 Ultimate Gedmatch calculator exactly measures? Does it measure ancient ancestry or more recent? What is the approximate period of time for MDLP K13 Ultimate test?

I do have some strange results, which don't resemble my known recent ancestry (Polish/Austrian) that's why I'm asking. How can I understand this results? My Eurogenes results look rather different. Welsh? English?


Primary Population (source) Secondary Population (source) Distance

1. 87.3% English_Cornwall_GBR + 12.7% Kurd 0.81
2. 87.4% English_Cornwall_GBR + 12.6% Uzbekistani_Jew 0.93
3. 87.3% English_Cornwall_GBR + 12.7% Kurd_WGA 0.93

Population (source) Distance
1 Welsh 3.4
2 North_European 3.58
3 Belgian 3.93
4 Slovenian2 4.07
5 German 4.37
1 Welsh 3.4
2 North_European 3.58
3 Belgian 3.93
4 Slovenian2 4.07
5 German 4.37
6 Germany_North 4.5
7 Germany_South 4.55
8 Slovak 4.68
9 North_German 4.7
10 Austrian 4.87
11 English_GBR 4.87

MDLP K13 Ultimate 4-Ancestors Oracle:

Using 1 population approximation:
1 Welsh @ 3.698860
2 North_European @ 3.829042
3 Slovenian2 @ 4.264181
4 Belgian @ 4.419270
5 German @ 4.453815
6 Germany_North @ 4.563188
7 North_German @ 4.717037
8 Slovak @ 4.827123
9 English_GBR @ 5.166817

Using 2 populations approximation:
1 50% France +50% Vepsa @ 2.673822

Using 3 populations approximation:
1 50% Icelandic +25% Spanish_Pais_Vasco_IBS +25% Tabasaran @ 1.789154


Thanks for any explanations!

[John McCoy]

One possible explanation is that the algorithm, whatever it is, was not tested on a sufficiently large sample to reveal that the results are sometimes way off the mark.

You should also be aware that claims about the "age" of the ethnic origins that are supposed to be detected have to be based on numerous assumptions, rather than actual data, because clearly nobody has gone back to, say, 1000 AD and sampled the ancestors of todays' various reference populations.

Perhaps the reason there are so many admixture algorithms is because there is not a consensus on how to measure this thing.

[loisrp]

It's basically saying - imagine a person who has the grandparents described. Based on the sample populations used by the calculator, you more or less (depending on the distance indicated by the number at the right) conform to the specified combination.

With one population, it's saying - the best we can do is say that you have four Welsh grandparents, but that's not a particularly good approximation.

With two populations, the best we can do is say you have 2 French grandparents and 2 Vespa (a particular Finnish population), and it's somewhat nearer than saying you are Welsh with 4 grandparents.

And so on.

With the first oracle, it's looking for the best combination of two populations. (The first is saying approximately 7 Cornish great-grandparents, and 1 Kurdish, and this oracle is not constrained by discrete blocks of grandparentage.)

As the combinations for this calculator are particularly fanciful, no one is expecting that you will indeed have the grandparents matching the descriptions. It's just meant to indicate the combinations that match the closest to the reality that is you.

[Armando]

Your example points out why I don't like Gedmatch very much. It has it's uses but I still like 23andme the most even with the higher cost and new experience problems.

[Dora]

John McCoy, loisrp, Armando,

Thank you so much for your great explanations.This kit actually belongs to my father, not me.

My thoughts are that this calculator probably approximates Austrians/Germans as French, and North Polish could maybe look close to Vepsa ( Finish population). It's definitely closer than English/Welsh.
However, maybe there is some truth to those results. My father's family was quite 'international', and I don't have the paper trail for two of my great-great grandparents, and also one great-grandmother was of some mixed heritage ( and no paper trail of her mother as well). My father has at least three Finish matches, French matches from Alsace ( which is very possible for my Austrian line) and also matches which seem to be English. Well, I don't know, maybe this calculator shows some little earlier, unknown ancestry of my father. I compared my father's results with my mom ( who is Polish, more Polish than my father) and she looks completely different in this calculator. It is still not right, but at least there are some Slavic populations in my mom's approximations, like Czech.

1 82.6% Sorbs + 17.4% Turk_Istanbul @ 0.97
2 85.8% Czech + 14.2% Cirkassian @ 1.29
3 85.9% Czech + 14.1% Kabardin @ 1.3

Using 2 populations approximation:
1 50% Don_cossack +50% Serbian_B-H @ 1.493707

Using 3 populations approximation:
1 50% European_Utah +25% Georgian +25% Russia @ 1.265350


My father should look like a mixture of Polish and Austrian (German), but there is almost nothing Slavic and the Austrian/German is quite far... ( which he shows in other calculators). Anyway, thank you for your great comments, I will investigate further. Maybe this calculator shows some older admixture. Thank you, again!

[loisrp]

There are very few people who come up exactly "right", particularly with MDLP. Maybe someone who is 99% Finnish.

[Dora]

Yes, of course. I don't expect it to be exactly right. Anyway, my mom's approximations look quite right