Gedmatch and G25 in many cases is a farce

[Lemgrant]

People often wondering why only 23andme shows accurate results, but everything else fails. Like for instance in case of Ljubic, a Bosnian-Croat with 12,5% German and some Bulgarian ancestry.
The reason his 12,5% German is undetectable by calculators, G25 and AncestryDNA, is because he has some more south influence, which overall places him back in his Bosnian-Croat average in terms of admixture proportions. While 23andme relies on IBD sharing info and accurately assigns NW euro to him. >> https://www.theapricity.com/forum/showthread.php?306878-My-23andMe-results

Here is example for Eurogenes k13. Let's say we have an average Austrian who is 12.5% Serbian and 12.5% English. So 75% Austrian, 12.5% Serbian, 12.5% English. As an Austrian, he has influence from North (English) and South (Serbian), which ends up in average Austrian results in admixture proportions.

https://i.imgur.com/IWNc5WY.png


k13(c(37.61,28.72625,14.6625,7.15125,8.7625,1.4937 5,0.76125,0.12875,0.22125,0.17125,0.2075,0.08125,0 .0125),k=30)
[,1] [,2]
[1,] "Austrian" "1.3376"
[2,] "East_German" "4.4093"
[3,] "Hungarian" "6.5641"
[4,] "West_German" "8.4429"
[5,] "South_Dutch" "9.7973"
[6,] "North_German" "11.5536"
[7,] "Croatian" "12.8212"
[8,] "Serbian" "12.9474"
[9,] "French" "13.5631"
[10,] "Danish" "14.4383"
[11,] "Moldavian" "14.4663"
[12,] "Swedish" "14.6663"
[13,] "North_Dutch" "14.7111"
[14,] "Southeast_English" "14.8363"
[15,] "North_Swedish" "15.5049"
[16,] "South_Polish" "15.7174"
[17,] "Orcadian" "16.0359"
[18,] "Norwegian" "16.1109"
[19,] "Southwest_English" "17.1136"
[20,] "Irish" "17.2369"
[21,] "Ukrainian_Lviv" "17.2987"
[22,] "Romanian" "17.4154"
[23,] "West_Scottish" "18.166"
[24,] "Ukrainian" "18.4077"
[25,] "Polish" "19.4827"
[26,] "Southwest_Finnish" "19.5934"
[27,] "PL_UKR_BEL_South_RUS" "20.7604"
[28,] "Bulgarian" "20.8938"
[29,] "Spanish_Cataluna" "21.1994"
[30,] "Spanish_Galicia" "21.3998"



k13(c(37.61,28.72625,14.6625,7.15125,8.7625,1.4937 5,0.76125,0.12875,0.22125,0.17125,0.2075,0.08125,0 .0125),k=30,mixedmode=T)
[,1] [,2]
[1,] "8.2% French + 91.8% Austrian" "0.5543"
[2,] "12.2% West_German + 87.8% Austrian" "0.6468"
[3,] "4.6% Spanish_Murcia + 95.4% Austrian" "0.6511"
[4,] "5.2% Spanish_Cataluna + 94.8% Austrian" "0.6707"
[5,] "4.5% Spanish_Valencia + 95.5% Austrian" "0.705"
[6,] "5% Portuguese + 95% Austrian" "0.7103"
[7,] "4.8% Spanish_Castilla_Y_Leon + 95.2% Austrian" "0.7157"
[8,] "4.1% Spanish_Andalucia + 95.9% Austrian" "0.737"
[9,] "4.5% Spanish_Extremadura + 95.5% Austrian" "0.7396"
[10,] "4.3% Spanish_Castilla_La_Mancha + 95.7% Austrian" "0.7503"
[11,] "3.9% Spanish_Aragon + 96.1% Austrian" "0.7522"
[12,] "4.7% Spanish_Galicia + 95.3% Austrian" "0.8083"
[13,] "9.8% South_Dutch + 90.2% Austrian" "0.8144"
[14,] "4.2% Spanish_Cantabria + 95.8% Austrian" "0.8477"
[15,] "4.2% North_Italian + 95.8% Austrian" "0.8608"
[16,] "3.3% Tuscan + 96.7% Austrian" "0.9385"
[17,] "3.9% Southwest_French + 96.1% Austrian" "0.9446"
[18,] "2.5% French_Basque + 97.5% Austrian" "0.9845"
[19,] "2.4% West_Sicilian + 97.6% Austrian" "1.034"
[20,] "2.5% Italian_Abruzzo + 97.5% Austrian" "1.0477"
[21,] "1.7% Sephardic_Jewish + 98.3% Austrian" "1.0581"
[22,] "1.7% Italian_Jewish + 98.3% Austrian" "1.0584"
[23,] "1.5% Tunisian_Jewish + 98.5% Austrian" "1.0905"
[24,] "1.9% South_Italian + 98.1% Austrian" "1.0941"
[25,] "1.5% Libyan_Jewish + 98.5% Austrian" "1.0987"
[26,] "1.6% Algerian_Jewish + 98.4% Austrian" "1.1048"
[27,] "95.2% Austrian + 4.8% Southeast_English" "1.1128"
[28,] "1.4% Sardinian + 98.6% Austrian" "1.1258"
[29,] "1.9% East_Sicilian + 98.1% Austrian" "1.1363"
[30,] "1.7% Ashkenazi + 98.3% Austrian" "1.1448"

[Ford]

Great post and 100% true. I know a half Italian and half Russian guy for example who would constantly be modelled as a pure Montenegrin.

[Samnium]

You have to take in account that you can inherit "almost" nothing from one great-grandparent

[Lemgrant]

You have to take in account that you can inherit "almost" nothing from one great-grandparent

Which is not the case for Ljubic and many others. IBD sharing clearly shows high sharing with Germans.

[Luke35]

The difference between 23adnMe and all the others is night and day for me. So, if IBD sharing is the major difference in their analysis method (versus the others), then bravo to them for adopting this approach.

[Zoro]

Well 23andme are not that innocent either. They also fudge to get people the desired results but they do it in a very smart way. Have you wondered why there is a sharp increase or dropoff in ancestry along their regional boundaries? For example people in Iran get let’s say 90% W. Asian and people in neighbor Afghanistan get 90% S. Asian. Do you really think those 2 peoples are that different from each other genetically? Of course not but here is the smart way of cheating.

According to them they pick a few 200 or 300 base pair windows in your genome and compare those windows to their references. Also according to them if the window is let’s say 60% Afghan instead of Iranian reference they make that window 100% S. Asian, and NOT 60% S. Asian. So they artificially purify it to 100% which in my book is fudging because if they had not fudged then people in Iran wouldn’t get 90% W. Asian and people in Afghanistan 90% S.Asian. Things would have been 60% or 50%.

Ofcourse they get away with it because chance or probability works in their favor because a guy with a window of 60% Afghan is more likely to be from Afghanistan than Iran therefore giving him 100% Afganistan or S. Asian works. Very smart!

[Jana]

Can I trust 23andme when it tells me I have lot of Slovenian ancestry?

[Lemgrant]

Well 23andme are not that innocent either. They also fudge to get people the desired results but they do it in a very smart way. Have you wondered why there is a sharp increase or dropoff in ancestry along their regional boundaries? For example people in Iran get let’s say 90% W. Asian and people in neighbor Afghanistan get 90% S. Asian. Do you really think those 2 peoples are that different from each other genetically? Of course not but here is the smart way of cheating.

According to them they pick a few 200 or 300 base pair windows in your genome and compare those windows to their references. Also according to them if the window is let’s say 60% Afghan instead of Iranian reference they make that window 100% S. Asian, and NOT 60% S. Asian. So they artificially purify it to 100% which in my book is fudging because if they had not fudged then people in Iran wouldn’t get 90% W. Asian and people in Afghanistan 90% S.Asian. Things would have been 60% or 50%.

Ofcourse they get away with it because chance or probability works in their favor because a guy with a window of 60% Afghan is more likely to be from Afghanistan than Iran therefore giving him 100% Afganistan or S. Asian works. Very smart!

they have Confidence Levels for this:

https://customercare.23andme.com/hc/en-us/articles/202906800-Confidence-thresholds-in-Ancestry-Composition

https://customercare.23andme.com/hc/en-us/articles/115004491928-Using-the-Advanced-Features-in-Ancestry-Composition


Confidence Levels

The algorithm we use to calculate your Ancestry Composition analyzes one small piece of your DNA at a time. For each piece of your DNA, we calculate the probability of that piece coming from 45 different populations. Each prediction is also linked to our confidence that the call is correct. By default, Ancestry Composition requires that our confidence in a prediction be greater than 50%. This confidence threshold allows the most DNA to be assigned and provides the most interesting view of your ancestry.

If you’d like to explore these assignments at a more advanced level, click on “Change confidence level” and move the slider that appears above your Chromosome Painting. The confidence slider on the Chromosome Painting allows you to explore our estimates of your genetic ancestry at different probability cutoffs. For example, if a segment of your DNA has a 55 percent chance of being Japanese, then that segment will be painted as Japanese at the 50 percent confidence level, but it will be painted with a more broad ancestry (either Broadly Japanese & Korean, Broadly East Asian & Native American, or Unassigned) at the 60 to 90 percent confidence levels.

Show us your results at 90% confidence level. The remaining "Broad" and "Unassigned" could be anything.

[Dick]

Only Ydna matters

[Lemgrant]

Can I trust 23andme when it tells me I have lot of Slovenian ancestry?

where do you see "a lot"? They don't give percentages of Ethnicities that are under the same category. For example mother gets Lithuanian under East Euro, but it is unknown how much % it is.

[Jana]

where do you see "a lot"? They don't give percentages of Ethnicities that are under the same category. For example mother gets Lithuanian under East Euro, but it is unknown how much % it is.

Croatian is not under Eastern Euro, and I get only Slovenian under it.

[Lemgrant]

Croatian is not under Eastern Euro, and I get only Slovenian under it.

ah yes, and you got Polish + Slovenian there before the update. If you get highest similarity with Slovakia in Global25 (which is between Croatia and Ukraine/Poland on PCA), it could indicate that you have some elevated northeast euro ancestry.

[Daos777]

g25 proves that

75 Ukrainian/Pole + 25 percent Moldavian = Slovakian haha

[Benyzero]

Only Ydna matters

Lol

[Lemgrant]

g25 proves that

75 Ukrainian/Pole + 25 percent Moldavian = Slovakian haha

There is really no difference between G25 and gedmatch eurogenes. G25 just has more samples. Ljubic is just average Bosnian with 0% German in G25, the same as on Gedmatch. Which is bullshit.

[Dick]

On a serious note, all companies should be using IBD since 2018.

[Ljubic]

Can I trust 23andme when it tells me I have lot of Slovenian ancestry?

Thats also a problem with 23andMe combining Greeks with Croats into one category which leads to using Eastern Euro for the former and Italian for the latter to average the results out. I agree that Gedmatch and G25 are interesting and have some merit. Gedmatchs oracles are only results from averages and should be taken with a grain of salt in my opinion but the admixture is the interesting part.
23andMe is more of a modern ancestry breakdown than the ancient one Gedmatch is

[Daos777]

I won’t believe any company until they accurately state my Scythian ancestry

[Lemgrant]

I won’t believe any company until they accurately state my Scythian ancestry

:icon_lol:

[Bakha]

I won’t believe any company until they accurately state my Scythian ancestry
Same shit here. I won’t believe any company until they detect my 50% Mongolian ancestry

[Lemgrant]

Same shit here. I won’t believe any company until they detect my 50% Mongolian ancestry

I know that you are joking, but that is not gonna happen. I saw your photos and I see nothing Mongolian in your appearance. Your central Asian side is mixed.

[JamesBond007]

People often wondering why only 23andme shows accurate results, but everything else fails. Like for instance in case of Ljubic, a Bosnian-Croat with 12,5% German and some Bulgarian ancestry.
The reason his 12,5% German is undetectable by calculators, G25 and AncestryDNA, is because he has some more south influence, which overall places him back in his Bosnian-Croat average in terms of admixture proportions. While 23andme relies on IBD sharing info and accurately assigns NW euro to him. >> https://www.theapricity.com/forum/showthread.php?306878-My-23andMe-results

Here is example for Eurogenes k13. Let's say we have an average Austrian who is 12.5% Serbian and 12.5% English. So 75% Austrian, 12.5% Serbian, 12.5% English. As an Austrian, he has influence from North (English) and South (Serbian), which ends up in average Austrian results in admixture proportions.

https://i.imgur.com/IWNc5WY.png


k13(c(37.61,28.72625,14.6625,7.15125,8.7625,1.4937 5,0.76125,0.12875,0.22125,0.17125,0.2075,0.08125,0 .0125),k=30)
[,1] [,2]
[1,] "Austrian" "1.3376"
[2,] "East_German" "4.4093"
[3,] "Hungarian" "6.5641"
[4,] "West_German" "8.4429"
[5,] "South_Dutch" "9.7973"
[6,] "North_German" "11.5536"
[7,] "Croatian" "12.8212"
[8,] "Serbian" "12.9474"
[9,] "French" "13.5631"
[10,] "Danish" "14.4383"
[11,] "Moldavian" "14.4663"
[12,] "Swedish" "14.6663"
[13,] "North_Dutch" "14.7111"
[14,] "Southeast_English" "14.8363"
[15,] "North_Swedish" "15.5049"
[16,] "South_Polish" "15.7174"
[17,] "Orcadian" "16.0359"
[18,] "Norwegian" "16.1109"
[19,] "Southwest_English" "17.1136"
[20,] "Irish" "17.2369"
[21,] "Ukrainian_Lviv" "17.2987"
[22,] "Romanian" "17.4154"
[23,] "West_Scottish" "18.166"
[24,] "Ukrainian" "18.4077"
[25,] "Polish" "19.4827"
[26,] "Southwest_Finnish" "19.5934"
[27,] "PL_UKR_BEL_South_RUS" "20.7604"
[28,] "Bulgarian" "20.8938"
[29,] "Spanish_Cataluna" "21.1994"
[30,] "Spanish_Galicia" "21.3998"



k13(c(37.61,28.72625,14.6625,7.15125,8.7625,1.4937 5,0.76125,0.12875,0.22125,0.17125,0.2075,0.08125,0 .0125),k=30,mixedmode=T)
[,1] [,2]
[1,] "8.2% French + 91.8% Austrian" "0.5543"
[2,] "12.2% West_German + 87.8% Austrian" "0.6468"
[3,] "4.6% Spanish_Murcia + 95.4% Austrian" "0.6511"
[4,] "5.2% Spanish_Cataluna + 94.8% Austrian" "0.6707"
[5,] "4.5% Spanish_Valencia + 95.5% Austrian" "0.705"
[6,] "5% Portuguese + 95% Austrian" "0.7103"
[7,] "4.8% Spanish_Castilla_Y_Leon + 95.2% Austrian" "0.7157"
[8,] "4.1% Spanish_Andalucia + 95.9% Austrian" "0.737"
[9,] "4.5% Spanish_Extremadura + 95.5% Austrian" "0.7396"
[10,] "4.3% Spanish_Castilla_La_Mancha + 95.7% Austrian" "0.7503"
[11,] "3.9% Spanish_Aragon + 96.1% Austrian" "0.7522"
[12,] "4.7% Spanish_Galicia + 95.3% Austrian" "0.8083"
[13,] "9.8% South_Dutch + 90.2% Austrian" "0.8144"
[14,] "4.2% Spanish_Cantabria + 95.8% Austrian" "0.8477"
[15,] "4.2% North_Italian + 95.8% Austrian" "0.8608"
[16,] "3.3% Tuscan + 96.7% Austrian" "0.9385"
[17,] "3.9% Southwest_French + 96.1% Austrian" "0.9446"
[18,] "2.5% French_Basque + 97.5% Austrian" "0.9845"
[19,] "2.4% West_Sicilian + 97.6% Austrian" "1.034"
[20,] "2.5% Italian_Abruzzo + 97.5% Austrian" "1.0477"
[21,] "1.7% Sephardic_Jewish + 98.3% Austrian" "1.0581"
[22,] "1.7% Italian_Jewish + 98.3% Austrian" "1.0584"
[23,] "1.5% Tunisian_Jewish + 98.5% Austrian" "1.0905"
[24,] "1.9% South_Italian + 98.1% Austrian" "1.0941"
[25,] "1.5% Libyan_Jewish + 98.5% Austrian" "1.0987"
[26,] "1.6% Algerian_Jewish + 98.4% Austrian" "1.1048"
[27,] "95.2% Austrian + 4.8% Southeast_English" "1.1128"
[28,] "1.4% Sardinian + 98.6% Austrian" "1.1258"
[29,] "1.9% East_Sicilian + 98.1% Austrian" "1.1363"
[30,] "1.7% Ashkenazi + 98.3% Austrian" "1.1448"

I see what you are saying and you make some good points but some people want to know what their number one closest modern population is. There are no pure Europeans for instance someone who is a mix of Irish, German and English may show up as English on non-ICB methods but this information is not false per se but is another interpretation I..E. it represents the person's most similar single modern ethnic group. Same for your Austrian, English and Croatian example and many other examples .

[Lemgrant]

I see what you are saying and you make some good points but some people want to know what their number one closest modern population is. There are no pure Europeans for instance someone who is a mix of Irish, German and English may show up as English on non-ICB methods but this information is not false per se but is another interpretation I..E. it represents the person's most similar single modern ethnic group. Same for your Austrian, English and Croatian example and many other examples .

Half Romanian and half Irish gets French or Austrian-Swiss as closest population. That information is useful only to see how evenly you got your Romanian and Irish genes from your parents. If you plot closer to Romanian on PCA (with Euclidean distances), that probably means that you are not exactly 50-50 of what you know.

[Bakha]

I know that you are joking, but that is not gonna happen. I saw your photos and I see nothing Mongolian in your appearance. Your central Asian side is mixed.
Yeah I know

[JamesBond007]

Half Romanian and half Irish gets French or Austrian-Swiss as closest population. That information is useful only to see how evenly you got your Romanian and Irish genes from your parents. If you plot closer to Romanian on PCA (with Euclidean distances), that probably means that you are not exactly 50-50 of what you know.

I dunno I think some Gedmatch calculators are more suited to different populations and that many suffer from the calculator effect. For instance, I think K15 is superior to K13 for the British isles because it breaks down the north-atlantic category into north sea and Atlantic categories but that it is missing some populations and that I would like to see on it and EUtest version 1 is the best for people like me who have no amerindian or east Asian admixture. Not all calculators are equal.

Someone who is 50 percent north European and 50 percent south European should better approximate French one would think.

[Grace O'Malley]

I agree that calculators will only take into account your overall genetics so if you are half English and half Serbian you will plot somewhere in Germany (approximately) but if 23&Me uses IBD sharing why do people like Italians get so much MENA for instance and why do North Africans get so much Iberian? Is that due to IBD. :confused:

[marco]

We don’t get so much Iberian in 23 and me lol who told you that. I’ve seen the latest updates for North Africans and we don’t score a lot of Iberian

[Grace O'Malley]

We don’t get so much Iberian in 23 and me lol who told you that

I've seen North Africans who do and also people like Armenians getting a lot of Italian and Greek. It's not uncommon from reading 23andMe.

[marco]

Show me these results in the latest update, I’m up to date on North African results and I haven’t seen any North Africans score. High percentage of Iberian in latest update

[Samnium]

I dunno I think some Gedmatch calculators are more suited to different populations and that many suffer from the calculator effect. For instance, I think K15 is superior to K13 for the British isles because it breaks down the north-atlantic category into north sea and Atlantic categories but that it is missing some populations and that I would like to see on it and EUtest version 1 is the best for people like me who have no amerindian or east Asian admixture. Not all calculators are equal.

Someone who is 50 percent north European and 50 percent south European should better approximate French one would think.

French average surely but actually the upper part of France can't be modeled like that.

The "french average" is drifted further south by the whole "southern" regions. SW french is iberian-like. SE France is somewhere between Iberia and N.Italy but a bit more northern. Britons and Normans are "british"-like. Here people drift strongly towards W.Germany.

I think the midpoint would be something like Central France.

[Lemgrant]

I dunno I think some Gedmatch calculators are more suited to different populations and that many suffer from the calculator effect. For instance, I think K15 is superior to K13 for the British isles because it breaks down the north-atlantic category into north sea and Atlantic categories but that it would s missing some populations and that I would like to see on it and EUtest version 1 is the best for people like me who have no amerindian or east Asian admixture. Not all calculators are equal.

Someone who is 50 percent north European and 50 percent south European should better approximate French one would think.

different samples were used in creation of various calculators.
http://dodecad.blogspot.com/2012/08/on-so-called-calculator-effect.html
http://bga101.blogspot.com/2012/05/beware-calculator-effect.html

all calculators use the same algorithm for their oracle:
2way mixedmode oracle was introduced by the Greek Dienekes >> http://dodecad.blogspot.com/2011/07/dodecad-oracle-v1.html
4-way oracle is by Alexandr Burnashev
and you can use nMonte algorithm for any gedmatch datasheet (just like in G25) https://anthrogenica.com/showthread.php?6452-How-to-Use-nMonte
you can also make custom mixedmode oracle with updated datasheet in RStudio >> https://www.theapricity.com/forum/showthread.php?177271-Tutorial-for-autosomal-calculators&p=6169153&viewfull=1#post6169153

The French and Swiss is just midpoint of Irish and Romanian on a PCA. So if you are 50-50 Irish-Romanian, you get fake similarity midpoint of your mix (closest population), which is French.

[marco]

You think ancestry will ever use IBD

[Pine]

You think ancestry will ever use IBD

They already do. IBD sharing isn't what distinguishes AncestryDNA from 23andMe. This whole rant can only distinguish 23andMe from calculators.

[Grace O'Malley]

Show me these results in the latest update, I’m up to date on North African results and I haven’t seen any North Africans score. High percentage of Iberian in latest update

It was very common. I remember reading on 23&Me forums about it and here is a quick search.

https://www.google.com/search?q=North+African+23andMe+results&sxsrf=ACYBGNQlaeherinYS1RjzrYkwTBGouOLmg:157529083 5652&source=lnms&tbm=isch&sa=X&ved=2ahUKEwjg57TV_5bmAhWEuZ4KHcGfC14Q_AUoAXoECA0QA w&biw=1366&bih=639#imgrc=b87DCBMxZuTYEM&imgdii=I8JCpWfazp7JIM

Whether it has been changed recently I don't know. It is just something that I recall. Italians also get a lot of Middle Eastern in their results.

The point though is that this surely can't be IBD?

[marco]

I’m unsure of before but currently with the updates that no longer happens

[marco]

What’s your thoughts on the caculators

[Grace O'Malley]

I’m unsure of before but currently with the updates that no longer happens

Well that's good to hear. But I'm still questioning the IBD if they are changing Ancestry's Comps so much. Someone might know more though. I know Ancestry must use IBD to a certain extent because of the Genetic Communities.

[JamesBond007]

different samples were used in creation of various calculators.
http://dodecad.blogspot.com/2012/08/on-so-called-calculator-effect.html
http://bga101.blogspot.com/2012/05/beware-calculator-effect.html

all calculators use the same algorithm for their oracle:
2way mixedmode oracle was introduced by the Greek Dienekes >> http://dodecad.blogspot.com/2011/07/dodecad-oracle-v1.html
4-way oracle is by Alexandr Burnashev
and you can use nMonte algorithm for any gedmatch datasheet (just like in G25) https://anthrogenica.com/showthread.php?6452-How-to-Use-nMonte
you can also make custom mixedmode oracle with updated datasheet in RStudio >> https://www.theapricity.com/forum/showthread.php?177271-Tutorial-for-autosomal-calculators&p=6169153&viewfull=1#post6169153

The French and Swiss is just midpoint of Irish and Romanian on a PCA. So if you are 50-50 Irish-Romanian, you get fake similarity midpoint of your mix (closest population), which is French.

Obviously, Davidski disagreed with Dienekes so we are not going to settle issue here. However, even barring that you are misinformed. Dodecad is biased i.e. it is Greco-Roman centric. MDLP is optimized for Lithuanians and Eurogenes K13 is meant for Euros with Asiatic admixture and K15 for Euros with Amerindians etc... admixture and EUtest version one is just meant for regular Europeans.Maybe, K13 hits the spot for most Euros because perhaps slavics outnumber westerners because the geography is bigger and poor people have more kids.

[Lemgrant]

They already do. IBD sharing isn't what distinguishes AncestryDNA from 23andMe. This whole rant can only distinguish 23andMe from calculators.


they do, yet it looks like if they don't, because their assignment is simply wrong in many cases. They gave 0% German to Ljubic. 0% French to LePrieur (with latest update, 1% before the update) https://www.theapricity.com/forum/showthread.php?278188-LePrieur-23andme-Results. Stears (Hungarian) got nothing Hungarian in his results. I (Ukrainian), got nothing Ukrainian in my results, they only show part of Ukraine together with Baltic States in latest update, before the update they were separate and nothing Ukrainian was showing for me, yet my top matches are all Ukrainians there. So what so difficult to show Ukraine on the map and conclude that I am not Lithuanian, or are we assimilated Balts with ukrainian names and surnames lol?

and many many other inaccurate results from AncestryDNA.

[marco]

they do, yet it looks like if they don't, because their assignment is simply wrong in many cases. They gave 0% German to Ljubic. 0% French to LePrieur (with latest update, 1% before the update) https://www.theapricity.com/forum/showthread.php?278188-LePrieur-23andme-Results. Stears (Hungarian) got nothing Hungarian in his results. I (Ukrainian), got nothing Ukrainian in my results, they only show part of Ukraine together with Baltic States in latest update, before the update they were separate and nothing Ukrainian was showing for me, yet my top matches are all Ukrainians there. So what so difficult to show Ukraine on the map and conclude that I am not Lithuanian, or are we assimilated Balts with ukrainian names and surnames lol?

and many many other inaccurate results from AncestryDNA.
I see it’s an interesting subject I might pick up a 23 and me test to see the difference

[Pine]

they do, yet it looks like if they don't, because their assignment is simply wrong in many cases. They gave 0% German to Ljubic. 0% French to LePrieur (with latest update, 1% before the update) https://www.theapricity.com/forum/showthread.php?278188-LePrieur-23andme-Results. Stears (Hungarian) got nothing Hungarian in his results. I (Ukrainian), got nothing Ukrainian in my results, they only show part of Ukraine together with Baltic States in latest update, before the update they were separate and nothing Ukrainian was showing for me, yet my top matches are all Ukrainians there. So what so difficult to show Ukraine on the map and conclude that I am not Lithuanian, or are we assimilated Balts with ukrainian names and surnames lol?

and many many other inaccurate results from AncestryDNA.

I made no statement about AncestryDNA's accuracy. When I have, I said that Ancestry is better a certain tasks, such as assigning Jewish ancestry to Western Jews when they do in fact have it.

[Luke35]

Half Romanian and half Irish gets French or Austrian-Swiss as closest population. That information is useful only to see how evenly you got your Romanian and Irish genes from your parents. If you plot closer to Romanian on PCA (with Euclidean distances), that probably means that you are not exactly 50-50 of what you know.

Of course that is assuming both of your parents are near the average for their ethnicity. Take my mother, who describes herself as of German ancestry when asked. It turns out, in fact, that she is closest to South Dutch, or somewhere between Belgian, Dutch, and West German on the K36 PCA. Wouldn't have known that if she had not tested, and this bit of info has helped me greatly in determining what I inherited from my dad.

[Kaspias]

Totally agreed.

[Lemgrant]

Of course that is assuming both of your parents are near the average for their ethnicity. Take my mother, who describes herself as of German ancestry when asked. It turns out, in fact, that she is closest to South Dutch, or somewhere between Belgian, Dutch, and West German on the K36 PCA. Wouldn't have known that if she had not tested, and this bit of info has helped me greatly in determining what I inherited from my dad.

and imagine that she also had some Hungarian ancestry that she wasn't aware of. She would plot like average German and would score like average German on gedmatch calculators, when in fact she would be a recent mix of 3 different ethnicities.

[Luke35]

and imagine that she also had some Hungarian ancestry that she wasn't aware of. She would plot like average German and would score like average German on gedmatch calculators, when in fact she would be a recent mix of 3 different ethnicities.

Exactly.

I just went thru a careful analysis of her ancestry, there is a lot of Bavarian there. However, Eurogenes says South Dutch. I guess her German mix, including the English she has, pulls her more North and West than expected.

[marco]

I just saw a girl on 23 and me score 92 percent North African which has clear admixture. So I wouldn’t call 23 and me accurate either, she also scored 100 percent Southern asian on Origuess and when we put her in group for rating everyone said indian or Pakistan.

[Lemgrant]

I just saw a girl on 23 and me score 92 percent North African which has clear admixture. So I wouldn’t call 23 and me accurate either, she also scored 100 percent Southern asian on Origuess and when we put her in group for rating everyone said indian or Pakistan.

wait, you judge her by phenotype and what she scores on origuess? Cumansky's mother scores North African on origuess and she is predominantly Ukrainian with possibly some Jewish. What is the other 8% at 50% confidence level results?. Also show photo.

[marco]

Duplicate glitched out

[Lemgrant]

I just saw a girl on 23 and me score 92 percent North African which has clear admixture. So I wouldn’t call 23 and me accurate either, she also scored 100 percent Southern asian on Origuess and when we put her in group for rating everyone said indian or Pakistan.

are you a bot? :icon_lol:
you posted the same thing with 33 minutes difference, or is it a forum glitch again?

my response is above

[marco]

are you a bot? :icon_lol:
you posted the same thing with 33 minutes difference, or is it a forum glitch again?

my response is above
Sorry bro glitched out and yes I showed her to my arab and North African group they stated the same thing she has clear admixture, 23 and me does not recognise this admixture. Sorry for late reply and I’ve seen this with many North African results where a lot of the components are being absorbed into NA. Her area has a lot of ssa admixture I was shocked to see none in her test I asked for her gedmatch she refused I would like to see how she scores on ancestry dna or Ftdna. She looks clear to have ssa admixture in her

[Lemgrant]

Sorry bro glitched out and yes I showed her to my arab and North African group they stated the same thing she has clear admixture, 23 and me does not recognise this admixture. Sorry for late reply and I’ve seen this with many North African results where a lot of the components are being absorbed into NA. Her area has a lot of ssa admixture I was shocked to see none in her test I asked for her gedmatch she refused I would like to see how she scores on ancestry dna or Ftdna. She looks clear to have ssa admixture in her
yes it’s true, 23andme does not show old ssa admixture for North Africans, it is embedded in their reference. For that you use gedmatch, she will get 10-20% ssa there.

[marco]

yes it’s true, 23andme does not show old ssa admixture for North Africans, it is embedded in their reference. For that you use gedmatch, she will get 10-20% ssa there.
I agree she refused to show her gedmatch she would score 20 percent ssa or above in those calculators, it’s very hard to determine ancestry based on single nucleotides because they are susceptible to mutations. I just see these tests as fun I don’t take them to seriously

[Lemgrant]

I agree she refused to show her gedmatch she would score 20 percent ssa or above in those calculators, it’s very hard to determine ancestry based on single nucleotides because they are susceptible to mutations. I just see these tests as fun I don’t take them to seriously

https://www.theapricity.com/forum/showthread.php?296847-Admixture-plots

[Lemgrant]

>>

[Lemgrant]

>>

[Lemgrant]

>>

[Peterski]

They do not use IBD for this part of the process:

"The algorithm we use to calculate your Ancestry Composition analyzes one small piece of your DNA at a time. For each piece of your DNA, we calculate the probability of that piece coming from 45 different populations."

My understanding is that they divide the whole DNA into segments about 300 SNPs long each, then they run ADMIXTURE calculator K45 separately on each segment, and then they assign the whole segment to its main component (for example if admixture run shows 55% West Asian 45% South Asian, they assign this segment to West Asian).

IBD is used only for assigning countries and regions, not for assigning major categories like East Euro or French-German. These are obviously done with ADMIXTURE or STRUCTURE.

The difference between 23andMe and AncestryDNA is that AncestryDNA runs ADMIXTURE on your whole DNA at once, without splitting it into segments and running on each segment.

IBD is used for Genetic Communities just like 23andMe is using it only for assigning countries and regions (without giving specific percentages).

GEDmatch is using it as well of course for family matching.

Davidski was also selling IBD analyses some time ago. Several users posted their IBD results from him - for example Grundig and Slavic Italian did.

[Lemgrant]

They do not use IBD for this part of the process:

"The algorithm we use to calculate your Ancestry Composition analyzes one small piece of your DNA at a time. For each piece of your DNA, we calculate the probability of that piece coming from 45 different populations."

My understanding is that they divide the whole DNA into segments about 300 SNPs long each, then they run ADMIXTURE calculator K45 separately on each segment, and then they assign the whole segment to its main component (for example if admixture run shows 55% West Asian 45% South Asian, they assign this segment to West Asian).

IBD is used only for assigning countries and regions, not for assigning major categories like East Euro or French-German. These are obviously done with ADMIXTURE or STRUCTURE.

The difference between 23andMe and AncestryDNA is that AncestryDNA runs ADMIXTURE on your whole DNA at once, without splitting it into segments and running on each segment.

IBD is used for Genetic Communities just like 23andMe is using it only for assigning countries and regions (without giving specific percentages).

GEDmatch is using it as well of course for family matching.

Davidski was also selling IBD analyses some time ago. Several users posted their IBD results from him - for example Grundig and Slavic Italian did.

they rely on IBD, otherwise they would assign SSA to North Africans for example, which is a clear result from ADMIXTURE run. But they score 0% ssa on 23andme.

[Peterski]

they rely on IBD, otherwise they would assign SSA to North Africans for example, which is a clear result from ADMIXTURE run.

Nope, it seems you misunderstood what I wrote.

If you run admixture on all SNPs at once, and if you do not use segment smoothing, you will get some % of SSA (for a North African person).

But they divide your DNA into hundreds of segments (200-300 SNPs each) and run it separately on each segment, and ignore minor components from each segment.

If your SSA admixture is not recent, then no segment will be majority SSA. For example segment "A" will be 20% SSA, segment "B" will be 34% SSA, segment "C" will be 14% SSA, and so on. All of these minor percentages are ignored, because they use "segment smoothing" (they assign the whole segment to its main component).

"Unassigned" percentages is the result of segments which did not have any predominant component (for example no component was over 50% on these segments).

=====

Old ancestries are fragmented, which means that even if someone is racially 20% Black, no segment will be majority SSA as long as this SSA is very old.

This is how North Africans who are 20% Black - but it is from a few thousand years ago - are able to score 100% North African and 0% SSA on 23andMe.

[Peterski]

The reason why some North Africans score 0% SSA on 23andMe is segment smoothing.

Let's say we have five segments (numbered 1. to 5.), each one with 200 markers (SNPs):

1. - 156 North African markers, 44 SSA markers = all segment, 100% of the segment (all 200 markers) assigned to North Africa
2. - 101 North African markers, 99 SSA markers = all segment assigned to North Africa
3. - 190 North African markers, 10 SSA markers = all segment assigned to North Africa
4. - 135 North African markers, 65 SSA markers = all segment assigned to North Africa
5. - 183 North African markers, 17 SSA markers = all segment assigned to North Africa

^^^
After segment smoothing, all five segments (all 1000 markers) will be assigned fully to North Africa.

The final result is, that this person scores 100% North African and 0% SSA (on these five segments).

If there are segments in which he has for example 120 SSA markers and 80 North African, then these segments are assigned Sub-Saharan, of course.

=====

And now why AncestryDNA is different? Because they do not use this segment smoothing method.

And probably they don't run admixture on each segment separately, but on all of the DNA at once.

Edit:

Segment smoothing is also why Mortimer was scoring just 2% South Asian on 23andMe at one point. His South Asian markers are very diluted in his DNA.

But after recent updates, it seems they have increased his South Asian to >10% again (which is surely more accurate for him, than that ridiculous 2%).

[Peterski]

They surely use IBD for assigning regions (which they started doing around the Spring or Summer of 2018).

But this method is not perfect either because it cannot tell the exact cause-effect relationship of a DNA match.

For example, I scored "Tyumen Oblast" among my regions and I'm 99% sure that I don't have any ancestors from there, on the other hand I'm also sure that some Poles who were sent to Siberia in the 19th century (as punishment for uprisings against Russia) ended up there.

So my match to Tyumen Oblast is most likely because of some Polish DNA there. Or, something ancient.

In general all Europeans share a lot of IBD with each other.

Obviously they show only IBD shared with nations counted as "East Euro" category under this category.

For my French-German, they show two regions in Germany as well as "Grand Est" and "Hauts de France".

These French regions might also be due to Polish migrants there (at least this is what mlukas suggested).

[Lemgrant]

The reason why some North Africans score 0% SSA on 23andMe is segment smoothing.

Let's say we have five segments, each one with 200 markers (SNPs):

1. - 156 North African markers, 44 SSA markers = all segment, 100% of the segment (all 200 markers) assigned to North Africa
2. - 101 North African markers, 99 SSA markers = all segment assigned to North Africa
3. - 190 North African markers, 10 SSA markers = all segment assigned to North Africa
4. - 135 North African markers, 65 SSA markers = all segment assigned to North Africa
5. - 183 North African markers, 17 SSA markers = all segment assigned to North Africa

^^^
After segment smoothing, all five segments (all 1000 markers) will be assigned fully to North Africa.

The final result is, that this person scores 100% North African and 0% SSA (on these five segments).

If there are segments in which he has for example 120 SSA markers and 80 North African, then these segments are assigned Sub-Saharan, of course.

=====

And now why AncestryDNA is different? Because they do not use this segment smoothing method.

And probably they don't run admixture on each segment separately, but on all of the DNA at once.

Edit:

Segment smoothing is also why Mortimer was scoring just 2% South Asian on 23andMe at one point. His South Asian markers are very dilluted in his DNA.

But after recent updates, it seems they have increased his South Asian to >10% again (which is surely more accurate for him, than that ridiculous 2%).

now this looks good, but this is the result on 50% confidence level. Also North Africans scored a lot European Iberian before the latest update on 23andme ~over 20% (user Adamm for example), but now 0%.

[marco]

The reason why some North Africans score 0% SSA on 23andMe is segment smoothing.

Let's say we have five segments (numbered 1. to 5.), each one with 200 markers (SNPs):

1. - 156 North African markers, 44 SSA markers = all segment, 100% of the segment (all 200 markers) assigned to North Africa
2. - 101 North African markers, 99 SSA markers = all segment assigned to North Africa
3. - 190 North African markers, 10 SSA markers = all segment assigned to North Africa
4. - 135 North African markers, 65 SSA markers = all segment assigned to North Africa
5. - 183 North African markers, 17 SSA markers = all segment assigned to North Africa

^^^
After segment smoothing, all five segments (all 1000 markers) will be assigned fully to North Africa.

The final result is, that this person scores 100% North African and 0% SSA (on these five segments).

If there are segments in which he has for example 120 SSA markers and 80 North African, then these segments are assigned Sub-Saharan, of course.

=====

And now why AncestryDNA is different? Because they do not use this segment smoothing method.

And probably they don't run admixture on each segment separately, but on all of the DNA at once.

Edit:

Segment smoothing is also why Mortimer was scoring just 2% South Asian on 23andMe at one point. His South Asian markers are very diluted in his DNA.

But after recent updates, it seems they have increased his South Asian to >10% again (which is surely more accurate for him, than that ridiculous 2%).
Ssa is clear in Some North Africans but didn’t see them score it in 23 and me very good break down and very disappointed to see this from 23 and me

[Peterski]

now this looks good

This is nothing new, Sikeliot started a thread about it here long ago and user Kurd (= Zoro?) on Anthrogenica:

https://www.theapricity.com/forum/showthread.php?208811-Does-23andme-overestimate-major-components-whilst-underestimating-minority-components


Also North Africans scored a lot European Iberian before the latest update on 23andme ~20% (user Adamm for example), but now 0%.

Adamm's European fell from 20% to 0% in the same update in which Mortimer's South Asian fell from 15% to 3%.

But it seems they later updated Mortimer's South Asian again upwards. Did Adamm's European stay at 0% ???

[Lemgrant]

This is nothing new, Sikeliot started a thread about it here long ago and user Kurd (= Zoro?) on Anthrogenica:

https://www.theapricity.com/forum/showthread.php?208811-Does-23andme-overestimate-major-components-whilst-underestimating-minority-components



Adamm's European fell from 20% to 0% in the same update in which Mortimer's South Asian fell from 15% to 3%.

But it seems they later updated Mortimer's South Asian again upwards. Did Adamm's European stay at 0% ???

no, it was 20% and fell to 0% in the latest update, while for Mortimer the south Asian score increased.

[Peterski]

no, it was 20% and fell to 0% in the latest update, while for Mortimer the south Asian score increased.

For Mortimer it first declined from 15% to 3% and then in another update it increased back from 3% to ca. 10%.

For Adamm it declined from 20% to 0% in the same update as for Mortimer, and later it probably stayed at 0%.

[Lemgrant]

For Mortimer it first declined from 15% to 3% and then in another update it increased back from 3% to ca. 10%.

For Adamm it declined from 20% to 0% in the same update as for Mortimer, and later it probably stayed at 0%.

his new results: https://www.theapricity.com/forum/showthread.php?301810-23andMe-beta-results-became-official-today-marking-the-end-of-the-beta-phase-testing

his previous results: https://www.theapricity.com/forum/showthread.php?296320-23andMe-results-came-in-today-(Berber-Moroccan)