G25 vs GEDmatch

[RyoHazuki]

Is there a consensus on what is superior for admixture vs PCA/plotting? I was under the impression G25 was best suited for plotting.

[gixajo]

Is there a consensus on what is superior for admixture vs PCA/plotting? I was under the impression G25 was best suited for plotting.

Mine is that G25 is best in both things, but both are useful to show what you are searching if you learn how to interprete them.

[Ion Basescul]

Both are as good as their references, but Global 25 is better for ancients. Global 25 is pretty weak for modern modelling of Eastern Europeans, compared to GEDmatch.

[J. Ketch]

GEDmatch is inferior in every way except for in modelling recent ancestry, in which case the updated Vahaduo gedmatch calcs are more accurate, but that's only because the modern samples are lacking in G25, Davidski won't add any that aren't academic (despite him admitting that modern Academic samples are often untrustworthy as well).

[vbnetkhio]

GEDmatch is inferior in every way except for in modelling recent ancestry, in which case the updated Vahaduo gedmatch calcs are more accurate,

eurogenes k13 works well for bronze age onwards as well, and gives very similar results to g25. it just can't make a good yamnaya/whg/ehg etc. model, but i'm not interested in that period, so i don't plan to ever buy G25.


but that's only because the modern samples are lacking in G25, Davidski won't add any that aren't academic (despite him admitting that modern Academic samples are often untrustworthy as well).

That's not true, he uses non-academic samples too, and refuses to add academics, when it suits him. The problem is, for ethnicites he dislikes, he deliberately makes averages that are 5-10% away from the truth. I have no other explanation for some of his averages.

He has been doing this since eurogenes k13 days. I was confused about South Slavic genetics for months until i realized how wrong his averages are.

[J. Ketch]

eurogenes k13 works well for bronze age onwards as well, and gives very similar results to g25. it just can't make a good yamnaya/whg/ehg etc. model, but i'm not interested in that period, so i don't plan to ever buy G25.



That's not true, he uses non-academic samples too, and refuses to add academics, when it suits him. The problem is, for ethnicites he dislikes, he deliberately makes averages that are 5-10% away from the truth. I have no other explanation for some of his averages.

He has been doing this since eurogenes k13 days. I was confused about South Slavic genetics for months until i realized how wrong his averages are.
Which ethnicities does he dislike (in your opinion)?

[Deusex99]

GEDmatch is inferior in every way except for in modelling recent ancestry, in which case the updated Vahaduo gedmatch calcs are more accurate, but that's only because the modern samples are lacking in G25, Davidski won't add any that aren't academic (despite him admitting that modern Academic samples are often untrustworthy as well).

There are plenty of non academic samples in G25. Unfortunately not all academic samples had their deep ancestry checked as well (like those Poles from Estonian research group with a lot of Siberian ancestry), so it's a mixed bag.

[vbnetkhio]

Which ethnicities does he dislike (in your opinion)?

South Slavs (there have been rumors, and the averages he made for them speak for themselves.) Also probably Italians and Greeks, since his arch-nemesis Dienekes(Dodecad) and Angela(Eupedia) are Greek and Italian.

It seems there is something fishy with his Italian averages as well https://www.theapricity.com/forum/showthread.php?326147-(((Davidski)))

[Rocinante]

There is no better thing than G25 in terms of modelling, this is FOR SURE.

[Zoro]

South Slavs (there have been rumors, and the averages he made for them speak for themselves.) Also probably Italians and Greeks, since his arch-nemesis Dienekes(Dodecad) and Angela(Eupedia) are Greek and Italian.

It seems there is something fishy with his Italian averages as well https://www.theapricity.com/forum/showthread.php?326147-(((Davidski)))


Agreed but G25 has bigger underlying issues. Simply put the admix proportions are not accurate ( not saying Gedmatch doesn’t have it’s problems).

Just a few things to show G25 is off.

1- SSA are closer to Europeans than some E Eurasians in G25. Even people here can understand this doesn’t make sense
2- Bedouin and some N Africans are modeled with significant Iran-Chl. Plain wrong. Iran-Chl didn’t have a significant impact west of Syria
3- N Africans are modeled with significant steppe. Not true in most cases. We can see this because most N Africans lack ANE

I can go on but don’t have time

[vbnetkhio]

Agreed but G25 has bigger underlying issues. Simply put the admix proportions are not accurate ( not saying Gedmatch doesn’t have it’s problems).

Just a few things to show G25 is off.

1- SSA are closer to Europeans than some E Eurasians in G25. Even people here can understand this doesn’t make sense
2- Bedouin and some N Africans are modeled with significant Iran-Chl. Plain wrong. Iran-Chl didn’t have a significant impact west of Syria
3- N Africans are modeled with significant steppe. Not true in most cases. We can see this because most N Africans lack ANE

I can go on but don’t have time
could this be fixed with a better choice of reference samples, or will a PCA always have these problems?

[Defcon2]

I can model myself with Guanche and North African samples obtaining up to 15%, I am going to request the complaint form from Davidski, this cannot be true.

[Zoro]

could this be fixed with a better choice of reference samples, or will a PCA always have these problems?

Not really it’s an inherent problem. PCA was never meant to do admix % and the way he runs it with uneven sample sizes in pops doesn’t help any

[Jana]

I prefer K13 over G25. G25 doesn't detect any western Euro in me. Either I really don't have any or it's false. I am open for either possibility.

[ph2ter]

G25 is better for ancients and it is not very good for moderns simply because it has too few modern regional averages.

I don't think that Davidski dislikes South Europeans, but rather he is only more interested in Poles and North Europeans and is fixated on R1a Y-DNA.

[Scandal]

How can I get G25

[Leto]

That's not true, he uses non-academic samples too, and refuses to add academics, when it suits him. The problem is, for ethnicites he dislikes, he deliberately makes averages that are 5-10% away from the truth. I have no other explanation for some of his averages.

He has been doing this since eurogenes k13 days. I was confused about South Slavic genetics for months until i realized how wrong his averages are.

South Slavs (there have been rumors, and the averages he made for them speak for themselves.) Also probably Italians and Greeks, since his arch-nemesis Dienekes(Dodecad) and Angela(Eupedia) are Greek and Italian.

It seems there is something fishy with his Italian averages as well https://www.theapricity.com/forum/showthread.php?326147-(((Davidski)))
At least he is not Russophobic enough to present Russians, East Slavs as quarter Siberian or something xD

[Defcon2]

G25 is better for ancients and it is not very good for moderns simply because it has too few modern regional averages.

I don't think that Davidski dislikes South Europeans, but rather he is only more interested in Poles and North Europeans and is fixated on R1a Y-DNA.

The French samples are horrible, but the Spanish ones are good although unequal in number of samples per region (in ancient and modern).

[gixajo]

The French samples are horrible, but the Spanish ones are good although unequal in number of samples per region (in ancient and modern).

The best Spanish reference is... Portuguese.:rolleyes:

Half of Spanish samples are from Cataluña, and are all differents, but well, it will improve as the number of references increases.

[Jana]

How can I get G25

Eurogenes Web Shop
https://bga101.blogspot.com/2017/10/genetic-ancestry-online-store-to-be.html

[Scandal]

I prefer K13 over G25. G25 doesn't detect any western Euro in me. Either I really don't have any or it's false. I am open for either possibility.

How much of your ancestry is western euro on paper?

[Cristiano viejo]

Both are as good as their references, but Global 25 is better for ancients. Global 25 is pretty weak for modern modelling of Eastern Europeans, compared to GEDmatch.

That is racisss

[Jana]

How much of your ancestry is western euro on paper?

Between 1/8 and 1/4, I estimated around 16% approximately. My mtdna is highly Germanic/NW Euro though and I did deep subclade test on it.
But my maternal grandma ancestry is very unclear, fact she was born in Vojvodina where lived all kinds of ethnics didn't help either. She refused to discuss her ethnic origins.

[gixajo]

The French samples are horrible, but the Spanish ones are good although unequal in number of samples per region (in ancient and modern).

I like French samples, I get similar distances as with Spanish ones.;)

Some accuse me of wanting to be French and of being Nordicist for doing that:rolleyes:

Although rather their intention is to avoid being considered as southern Europeans? :confused:

[Defcon2]

I like French samples, I get similar distances as with Spanish ones.;)

Some accuse me of wanting to be French and of being Nordicist for doing that:rolleyes:

Although rather their intention is to avoid being considered as southern Europeans? :confused:

If the French are southern Europeans, how come they accuse you of that :D

[Dr_Maul]

For me, G25 is a lot mroe accurate distance wise

[Scandal]

Between 1/8 and 1/4, I estimated around 16% approximately. My mtdna is highly Germanic/NW Euro though and I did deep subclade test on it.
But my maternal grandma ancestry is very unclear, fact she was born in Vojvodina where lived all kinds of ethnics didn't help either. She refused to discuss her ethnic origins.

I've just checked your k13 vahaduo, you do seem partially western. I too seem to be somewhat western admixed on the same calc("swiss_italian"), or is it just native balkan admixture? I'm trying to make sense of my result.

[ph2ter]

The French samples are horrible, but the Spanish ones are good although unequal in number of samples per region (in ancient and modern).

Some French areas really are not good enough. For example Pas-de-Calais and Seine-Maritime do not look very representative to me when I make my maps.
Then, Italian_Liguria is only one sample...

[RyoHazuki]

I like French samples, I get similar distances as with Spanish ones.;)

Some accuse me of wanting to be French and of being Nordicist for doing that:rolleyes:

Although rather their intention is to avoid being considered as southern Europeans? :confused:

That is absurd. Now France is accepted Nordicist group? They gerrymander who is/isn't Nordicist material depending on when it's convenient. France does something they like? Nordic. Otherwise? Just smelly gallo romans.

[Ion Basescul]

That is racisss

The guy from Eurogenes wants to include only academic samples in his database for Global 25, but there are few studies with modern samples available from Eastern Europe. On GEDmatch though, there are ton of people.

[vbnetkhio]

The guy from Eurogenes wants to include only academic samples in his database for Global 25, but there are few studies with modern samples available from Eastern Europe. On GEDmatch though, there are ton of people.

Eastern Europe actually has better academic sample coverage than Western, Davidski just refuses to add all the available samples. when he sees it fit, he adds non-Academics too.

[vbnetkhio]

G25 is better for ancients and it is not very good for moderns simply because it has too few modern regional averages.

I don't think that Davidski dislikes South Europeans, but rather he is only more interested in Poles and North Europeans and is fixated on R1a Y-DNA.

i can't know his exact motivation, put he definitely puts an effort into "sabotaging" his own averages.

[Ion Basescul]

Eastern Europe actually has better academic sample coverage than Western, Davidski just refuses to add all the available samples. when he sees it fit, he adds non-Academics too.

The countries that interest me (Moldova, Romania, Western Ukraine, Bulgaria) don't, as far as I'm aware.

[Jana]

I've just checked your k13 vahaduo, you do seem partially western. I too seem to be somewhat western admixed on the same calc("swiss_italian"), or is it just native balkan admixture? I'm trying to make sense of my result.

It can be, yeah. Native Balkanites from Croatia, Hungary and Moldova were partly western Euro (celtic like)
But we don't know amount of Roman impact in balkans and panonnia. Roman admixture would pull native Balkanites towards east mediterranean, so it's hard to know what is ancient and what is "recent" west euro like admix.

[JerryS.]

South Slavs (there have been rumors, and the averages he made for them speak for themselves.) Also probably Italians and Greeks, since his arch-nemesis Dienekes(Dodecad) and Angela(Eupedia) are Greek and Italian.

It seems there is something fishy with his Italian averages as well https://www.theapricity.com/forum/showthread.php?326147-(((Davidski)))

I have run my Ancestry and 23 data through the various GEDmatch models as well as using the scaled and unscaled raw data G25 with vahaduo. I've said before that I think Eurogenes and G25 has a northern lean or bias based on my oracle results, while Dodecad V3 and MDLP K23b seem more accurate for me based on how they reflect my minor Mediterranean ancestry. but this is just my layman view.

[vbnetkhio]

The countries that interest me (Moldova, Romania, Western Ukraine, Bulgaria) don't, as far as I'm aware.

there are no Irish and Danish acamedic samples available. not a single one.

compare it to this: for Moldova there are those older, mixed samples from 2 different studies i uploaded to gedmatch, and there are those newer, south plotting ones from southern Moldova. there are also 2 sets of Gagauz samples.
for Romania there are 3 sets of samples, Gorj, Apuseni and those labeled just as Romanian.
for Bulgaria there are 2 sets of samples, one from Plovdiv and the other labeled just as Bulgarian.
for Western Ukraine there are 7 samples from Lviv, one or 2 from Chernivtsi, 4 from Zhytomir, 3 from Rivne and a Hungarian from Transcarpathia.

[Leto]

there are no Irish and Danish acamedic samples available. not a single one.

What was used for Eurogenes then?

[vbnetkhio]

What was used for Eurogenes then?
results of Davidski's buddies. or maybe some non-public samples that Davidski has access to.

[Ion Basescul]

there are no Irish and Danish acamedic samples available. not a single one.

compare it to this: for Moldova there are those older, mixed samples from 2 different studies i uploaded to gedmatch, and there are those newer, south plotting ones from southern Moldova. there are also 2 sets of Gagauz samples.
for Romania there are 3 sets of samples, Gorj, Apuseni and those labeled just as Romanian.
for Bulgaria there are 2 sets of samples, one from Plovdiv and the other labeled just as Bulgarian.
for Western Ukraine there are 7 samples from Lviv, one or 2 from Chernivtsi, 4 from Zhytomir, 3 from Rivne and a Hungarian from Transcarpathia.

Well, yes, but not all of them made it Global 25. For example I'm aware of only one sample from Western Ukraine there, the one that's labelled Lviv.
The Romanian samples come from around the same area. Gorj and Alba counties are pretty much next door.
The ones from Moldova are pretty good though, as we have a dataset from the North of the country and one from the South. The ones from the third Eurogenes dataset are pretty much Central looking between the two data sets, if we were to exclude the full Russians/Ukrainians.
For Bulgaria there are only 5 samples in the spreadsheet.

He has about 68 Irish coordinates though, so I mean that's still better than having 5 academic samples if you ask me. And I remember someone saying that he's added Danes too, for which I'm not sure if they are sourced from studies.

[vbnetkhio]

Well, yes, but not all of them made it Global 25. For example I'm aware of only one sample from Western Ukraine there, the one that's labelled Lviv.


because Davidski has some issues.



The Romanian samples come from around the same area. Gorj and Alba counties are pretty much next door.
The ones from Moldova are pretty good though, as we have a dataset from the North of the country and one from the South. The ones from the third Eurogenes dataset are pretty much Central looking between the two data sets, if we were to exclude the full Russians/Ukrainians.

But one is in Wallachia and the other in Transylvania. Maybe the sampling points could've been better, but it's better than having samples form one region only, or no samples at all as it is for some other countries.



For Bulgaria there are only 5 samples in the spreadsheet.
He has about 68 Irish coordinates though, so I mean that's still better than having 5 academic samples if you ask me. And I remember someone saying that he's added Danes too, for which I'm not sure if they are sourced from studies.

he literally just took the first 5 samples (alphabetically) from the Reich lab dataset , and used just those. and considers this to be representative. who knows how they plot compared to the actual average.

this means he went out of his way, opened the Reich dataset, sorted the samples alphabetically by name, and manually removed all samples except the first 5.
he did the same for Bosniaks, Montenegrins, Serbs and Macedonians. (and for all other ethnicites too, but there the "threshold" wasn't 5, but 10 or more)

He could have just been lazy and kept all the academic samples, but he put in extra effort just to make an unrepresentative sample choice and shitty averages.
and to think i almost paid him for this mess of a tool.

[Ion Basescul]

because Davidski has some issues.



But one is in Wallachia and the other in Transylvania. Maybe the sampling points could've been better, but it's better than having samples form one region only, or no samples at all as it is for some other countries.



Definitely, but being on the fringes of Wallachia, they are very similar genetically to the people from Alba. Although phenotypically, they look quite different IMO.

https://i.imgur.com/H1RZLCi.png

https://i.imgur.com/a7Mi60N.png

The exact location for the Gorj samples is Tismana (https://ro.wikipedia.org/wiki/Tismana) village by the way.
These are the people from there:

https://www.youtube.com/watch?v=TKY8gypIvLo

For Apuseni, all samples come from Horea (https://ro.wikipedia.org/wiki/Horea,_Alba) village in Alba county, Transylvania.

https://www.youtube.com/watch?v=gIJEfF7XzVA