PDA

View Full Version : MyHeritage and inconsistent Ancestry Assignments - My Impression



Riverman
07-22-2020, 10:47 AM
I want to open up a thread about how MyHeritage seems to work, which is good and what is bad about its algorithm and why some results seem to be clearly inconsistent.

Basically I think MyHeritage does a decent job with its ancestry assignment, but overall its still behind the leader in this field to me, which is 23andme. And the reason for its strength and weakness seems to be that they use a similar method as does Oracle or similar calculators especially if its about undersampled populations and subpopulations. Which means their algorithm takes what is closest in a classic admixture run, but without checking actual identity by descent. To explain that further, it might mean that some individuals get something like "Skandinavian + West Asian", while others get, for the same shared segments, "Balkan" if coming from an undersampled group without proper references.
This is a mistake 23andme rarely does, their assignment seems to be closer to real descent, rather than a simple admixture prediction. And if they lack the reference completely, they skip it as "undeclared" or a general category.

I checked that in detail for clusters of closely related, fairly endogamous people. Like groups which share a minor Balkan-related ancestry. Yet while this is clearly defined on 23andme and swallowed on FTDNA, LivingDNA etc. by one single major European component, on MyHeritage it gets assigned randomly. Like some people get Scandinavian:West Asian, others get Balkan:Middle Eastern, again others Eastern European:West Asian, the next one Greek, another Italian or even Iberian. So the MH algorithm tries to make sense out of this minor ancestral, rather South Eastern component, but doesn't have the right references and can't make sense out of the mixture, so it assigns it to very different extreme or less extreme admixture ratios, with the only consistency being "Southern/South Eastern ancestry" in the broadest thinkable way.

This is not the case, by the way, for more clearly defined, large ancestral components and those with a very good reference data base and from a signle ethnicity, even with a similar basic ancestry. The algorithm recognises them as a regional group and assigns them, largely, correctly. But for those which are more mixed and with a bad reference for their real ancestry, the algorithm seems to become erratic and random in the way described. I'm pretty sure other people with a mixed background and no good reference data experienced the same on MyHeritage.

In this respect the result is inferior to 23andme, but on a similar level as FTDNA (which is a swallower however). I can also confirm what others have said in the MH results thread: FTDNA pulls people East, MH pulls people South East and Living DNA is the big swallower for Central Europeans - probably better for British people. 23andme has the problem of undeclared, but what they declare even at the speculative level makes the most sense from my experience, not just for me, but for different clusters of relatives from different ethnic groups in Europe.

23abc
07-22-2020, 11:06 AM
I think I've mentioned this in other posts before, but MyHeritage definitely works something more similar to an oracle than anything else. It tries to find an optimal fit, rather than identify actual admixture, which can lead to people of the same ethnic background scoring in completely different categories. For example, some of my matches lose most of their 'Greek and Southern Italian' for a combination of Sardinian and Sephardic Jew instead. Obviously, these same people score 0% Sardinian on 23andme, because they aren't actually Sardinian. I've actually never experienced a good result on MyHeritage for any of the kits I've purchased for myself and family, which sits at 4 currently.

In my opinion, what matters most for the admixture test is if it's consistent between people of the same ethnic background. If everyone with similar ethnic background scores in the same categories, with similar levels of %, then I say the test is good. 23andme does this well. Even if I score mostly Italian instead of Greek, so do all of my matches of same ethnic background - the levels of Italian and WANA are very consistent for all Dodecanese islanders. MyHeritage on the other hand, has my matches in completely different categories in very different %. So I think its quality is poor.

I'm waiting on my AncestryDNA to rate that one, but it seems to have some issues also. We can only hope that MyHeritage's update is not just increasing the amount of regions, but actually fixing the consistency of their admixture results.

Ibericus
07-22-2020, 11:10 AM
Yes, MH is just a calculator. They don't seem to look at your matches when assigning ancestry. As an example they make me 20% Italian and I only have 5 Italian matches (out of 900).

On the other and 23andme is too conservative for my taste. In my case only 67% of my matches have Spanish ancestry and yet they seem to believe that my ancestry is pretty average beacuse I don't get anything 'exotic'.

geebee
07-22-2020, 11:56 AM
I think one of the strengths of 23andMe as opposed to its competitors -- including MyHeritage -- is that they don't just give percentages for any ancestry. They actually show each ancestral component at its location in the genome. This permits direct comparisons, which percentages alone do not.

Even if you see that you and a cousin both have x percent of some ancestry, there's no way to tell if your shared matches reflect that ancestry. (Unless, of course, you're both 100% of the same thing.)

But FYI, even 23andMe sometimes gets it wrong. My daughter has a segment on chromosome 7 that is identified as "Spanish & Portuguese". It's part of a slightly longer segment identified as "Southern European", but the very end of this segment is simply "Broadly Southern European".

This segment is shown as paternal for my daughter -- from me. Only, I don't have a "Spanish & Portuguese" segment here (according to 23andMe). In fact, I don't even have a "Southern European" segment. Instead I have a "Broadly Northwestern European" segment on my mother's side (which is the side that includes Spanish ancestry), and a "French & German" segment on my father's side.

As it happens, my daughter does not match my father on any part of her paternal copy of chromosome 7, meaning she appears to have inherited an unrecombined copy of my maternal chromosome 7. My mother was one-fourth Spanish, and her mother was half Spanish. However, this was not because either of my grandmother's parents was Spanish, but because both of them were also half Spanish. (Both my grandmother's grandfathers were Spanish immigrants.)

In other words, it would be quite reasonable for my daughter's Spanish segment on chromosome 7 to be real, but why isn't the same segment shown as Spanish for me? The only reason I can figure is that her Ancestry Composition is more accurate than mine. This would make sense, since my daughter has two tested parents and I only have one (my father).

But my point is simply that all testing companies sometimes make errors in their assignments. It's just harder to tell when the only thing you have to look at is percentages. Unfortunately, this is the case for MyHeritage. Until they show ancestry in graphic form rather than just by percents, it will continue to be a problem in any comparison.

Riverman
07-22-2020, 12:24 PM
But my point is simply that all testing companies sometimes make errors in their assignments. It's just harder to tell when the only thing you have to look at is percentages. Unfortunately, this is the case for MyHeritage. Until they show ancestry in graphic form rather than just by percents, it will continue to be a problem in any comparison.

23andme seems to use an algorithm which assigns whole segments to a specific category. I don't know the exact size any more, but there is a specific rule to it. This means that a segment which is actually 60:40 - NW:SW would be assigned NW. But probably if the same segment is 50:30:20 for SW:NW:SE, it will assign to SW again. This means to me that, probably, that segment of your daughter is probably not more Spanish, but less North Western than for you. So while it may be just a lapse of 23andme, it could instead mean that the segment changed between the generations only slightly, or the boundaries have changed, resulting in a different assignment.

The direct comparison 23andme allows is a great asset they offer indeed, so you can check things yourself. Yet they just make less mistakes in general, even with the percentages. Like most other admixture proportions provided by other companies are just way off for me, as they are for other members of related clusters, but 23andme just nailed it in comparison. They are just not perfect where they lack references, but their algorithm has not such a big problem with mixtures generally speaking. MyHeritage really struggles not just with undersampled, but also ethnically mixed people in general is my impression.

And I completely agree with 23abc: Consistency is in the title of the thread, because that's what it is about. If you have 10 related people from one ethnic group, from one small region, with surely largely the same ancestry proportions, you can at least expect consistent results. 23andme does deliver something better, even if still imperfect. But they largely vary on the sub-level, so below the major divisions of North Western European and Southern European. I selected a fairly homogeneous group of Southern Germans and Western Austrians for a comparison:
In this group of people the NWE and SE stays pretty constant overall (single exception for SE), what changes is primarily the assignment to the subdivisions. To give an example from a homogeneous 5 sample group for the ranges:
NWE: 81,1-83,9 (= 2,8)
F&G: 58,8-68,1 (= 9,3)
B&I: 2,1-5,2 (= 3,1)
Scandi: 0-1,5 (= 1,5)
BNWE: 8,3-18,9 (= 10,6)

Notice the only slight difference in general NWE and the large discrepancies between the subdivisions. This is not perfect, but if I compare similar results from MyHeritage, its still far more consistent, like they all get British & Irish, as wrong as it is. You know Southern Germans get it, its actually just German/Celto-Germanic, wrongly assigned because of the much larger and better constructed British & Irish reference, many possibly mixed American samples in the reference. But its not completely off, like something "more exotic" thrown in to balance 20 percent Scandinavian out.

lacreme
07-22-2020, 12:53 PM
I think I've mentioned this in other posts before, but MyHeritage definitely works something more similar to an oracle than anything else. It tries to find an optimal fit, rather than identify actual admixture, which can lead to people of the same ethnic background scoring in completely different categories. For example, some of my matches lose most of their 'Greek and Southern Italian' for a combination of Sardinian and Sephardic Jew instead. Obviously, these same people score 0% Sardinian on 23andme, because they aren't actually Sardinian. I've actually never experienced a good result on MyHeritage for any of the kits I've purchased for myself and family, which sits at 4 currently.

In my opinion, what matters most for the admixture test is if it's consistent between people of the same ethnic background. If everyone with similar ethnic background scores in the same categories, with similar levels of %, then I say the test is good. 23andme does this well. Even if I score mostly Italian instead of Greek, so do all of my matches of same ethnic background - the levels of Italian and WANA are very consistent for all Dodecanese islanders. MyHeritage on the other hand, has my matches in completely different categories in very different %. So I think its quality is poor.

I'm waiting on my AncestryDNA to rate that one, but it seems to have some issues also. We can only hope that MyHeritage's update is not just increasing the amount of regions, but actually fixing the consistency of their admixture results.


Looks like that way indeed
My friend's results...
A reminder that he is Greek with ancestry from multiple and different parts of the country,the detailed breakdown of which is available on other posts
https://i.imgur.com/m3qvVYy.jpg

Riverman
07-22-2020, 01:00 PM
Looks like that way indeed
My friend's results...


Did he test with another company too? If so, how did he score there?

lacreme
07-22-2020, 01:23 PM
Did he test with another company too? If so, how did he score there?

Not yet
But he has transfered his raw dna to ftdna and livingdna and he gets the following results

https://i.imgur.com/IsHbHMS.jpg
https://i.imgur.com/FllhoRK.jpg

Riverman
07-22-2020, 01:47 PM
Transferred data is as good imho, but unfortunately 23andme doesn't accept it afaik.

Caius Agrippa
07-22-2020, 01:56 PM
Any test that does not look at actual matches and IBD to assign ancestry will come out with relatively innacurate ancestry predictions for most people. My results on MyHeritage actually make sense if I interpret them the way you describe, but they are innacurate as a ''literal picture''. Just for example, I have some ancestry from an extremely undersampled part of Italy (Trentino/Friuli) and they likely make my Northeast Italian ancestry a mix of Italian, West Asian, Iberian and Eastern European.

Riverman
07-22-2020, 02:16 PM
Any test that does not look at actual matches and IBD to assign ancestry will come out with relatively innacurate ancestry predictions for most people. My results on MyHeritage actually make sense if I interpret them the way you describe, but they are innacurate as a ''literal picture''. Just for example, I have some ancestry from an extremely undersampled part of Italy (Trentino/Friuli) and they likely make my Northeast Italian ancestry a mix of Italian, West Asian, Iberian and Eastern European.

Its not just about some components appearing, but also the size of them. Their "ethnicity map" tool is helpful in checking this, since you can look at how many customers from a given country show a specific ancestral component. Actual Italian customers get the component "Italian" at 76,1 % (doesn't tell you how much of it, but seems rather low, even if considering non-Italians residing and buying from Italy) and "Iberian" at 41,1 %. "Balkan" and "Eastern European" is very, very close and can switch in individuals.

Caius Agrippa
07-22-2020, 07:10 PM
Its not just about some components appearing, but also the size of them. Their "ethnicity map" tool is helpful in checking this, since you can look at how many customers from a given country show a specific ancestral component. Actual Italian customers get the component "Italian" at 76,1 % (doesn't tell you how much of it, but seems rather low, even if considering non-Italians residing and buying from Italy) and "Iberian" at 41,1 %. "Balkan" and "Eastern European" is very, very close and can switch in individuals.

It seems some Italian ethnic groups don't even score Italian on these DNA tests. People from North Italy score very low amounts of Italian. I have seen some with even more Balkan or Iberian than Italian.

Riverman
07-23-2020, 08:46 AM
It seems some Italian ethnic groups don't even score Italian on these DNA tests. People from North Italy score very low amounts of Italian. I have seen some with even more Balkan or Iberian than Italian.

Its explainable by the population history of the region, but its still a bad algorithm. But even worse is really the inconsistency, because if I use the same Southern German and Western Austrian regional samples for a comparison, they get fairly regular, meaningful results on 23andme, while MyHeritage and others really just pick randomly ancestral components to construct a synthetic approximation. One gets more Scandinavian+British and more Balkan or even West Asian and Jewish, the next one gets almost no Scandinavian and a lot of Italian-Greek to come to the same position on a PCA. But it has zero to do with the real recent ancestry (last 1000 years), its just an admixture run without checking the identity of the segments, what 23andme does and its all apparently randomly chosen, even for family members.
There are even Bavarian-Austrian people which score zero (!) North and West European, while their cousins are something like 80-90 percent because of this splits. They usually get English or Scandinavian instead, plus something very Southern to South Eastern to balance this out. Really a bad algorithm.

Claudio
07-23-2020, 08:55 AM
I think I've mentioned this in other posts before, but MyHeritage definitely works something more similar to an oracle than anything else. It tries to find an optimal fit, rather than identify actual admixture, which can lead to people of the same ethnic background scoring in completely different categories. For example, some of my matches lose most of their 'Greek and Southern Italian' for a combination of Sardinian and Sephardic Jew instead. Obviously, these same people score 0% Sardinian on 23andme, because they aren't actually Sardinian. I've actually never experienced a good result on MyHeritage for any of the kits I've purchased for myself and family, which sits at 4 currently.

In my opinion, what matters most for the admixture test is if it's consistent between people of the same ethnic background. If everyone with similar ethnic background scores in the same categories, with similar levels of %, then I say the test is good. 23andme does this well. Even if I score mostly Italian instead of Greek, so do all of my matches of same ethnic background - the levels of Italian and WANA are very consistent for all Dodecanese islanders. MyHeritage on the other hand, has my matches in completely different categories in very different %. So I think its quality is poor.

I'm waiting on my AncestryDNA to rate that one, but it seems to have some issues also. We can only hope that MyHeritage's update is not just increasing the amount of regions, but actually fixing the consistency of their admixture results.

Have you checked through your DNA matches for Eastern Sephardic Matches?
It's always possible you could have partial real Sephardic ancestry?

Riverman
07-23-2020, 09:25 AM
Have you checked through your DNA matches for Eastern Sephardic Matches?
It's always possible you could have partial real Sephardic ancestry?

I have seen people with real Jewish or Near Eastern ancestry and without on MyHeritage. Its being almost randomly assigned in the way described, oftentimes just as a minor admixture component to balance out a too North Western ancestral component for Central Europeans (like English, Irish, Scandinavian instead of general Northern and Western European). I could check it for cousins from the same clusters, one gets it, the other gets the more appropriate NW:Balkan:South Italian mix, which is just another admixture experiment and not identity too however. None of them has Jewish/Near Eastern on 23andme above trace levels and they are all fairly North Western in Oracle and on the PCA. So while its always possible the admixture is real, a hit on MyHeritage is the least reliable starting point.

23abc
07-23-2020, 09:35 AM
Have you checked through your DNA matches for Eastern Sephardic Matches?
It's always possible you could have partial real Sephardic ancestry?

Filtering by Sephardic Jew or Ashekanzi only has Greek people on MyHeritage. And it's oracle-like, because the same thing occurs on oracles (K13, G25, etc.) for many Greeks if Jewish populations are included. The high amounts of Sardinian is also good evidence, as this is a unique admixture that only appears in oracles, not tests that actually detect patterns in segments like 23andme.

mildlycurly
07-23-2020, 01:43 PM
@Riverman, your explanation makes a lot of sense in light of what I've seen.

Many of my matches have small amounts of what appears to be Romani ancestry based on their family trees. Yet they tend to score Balkan or Middle Eastern instead of South Asian. This is because the Romani are already so mixed that the algorithm can't exactly place it. It's a similar situation with oracles and nMonte- the combined northern MENA and South Asian influence tends to show up as various Caucasian ethnicities.

As an interesting aside, my paternal Romani on 23andme shows up as Broadly Western Asian whereas my maternal Romani is South Asian! So while 23andme is more accurate than MH in the long run, it can still make mistakes.

Riverman
07-23-2020, 01:56 PM
As an interesting aside, my paternal Romani on 23andme shows up as Broadly Western Asian whereas my maternal Romani is South Asian! So while 23andme is more accurate than MH in the long run, it can still make mistakes.

Yes, they are for sure not perfect and make mistakes. Also, like so often in life, there is sometimes no one and only best solution, but just a best compromise available. But looking at your example in detail, do you mean the same segment or just the same presumed ancestry, one time from your fathers, one time from your mothers side? Because Romani ancestry is in itself, like you said, fairly mixed, so they can have actual South Asian, but also West Asian and Balkan, even Jewish or Eastern European and so on. I also noticed differences in Roma people both in phenotype, as well as in genotype, from different regions like Slovakia vs. Serbia. One study, I don't have the link at hand, showed that too, with different non-South Asian admixture proportions and, especially paternal, haplogroups.

Dr_McNinja
07-23-2020, 04:09 PM
It definitely feels like an Oracle result in terms of how inconsistent it is. At least with the Oracles, they give you several different models which helps one understand what you're seeing (it's just modeling).

You can use Oracles for ancestry composition, but it would require a different set of limited/narrowed-down populations for each region or ethnic group.

I made a pretty accurate one (using a Gedmatch calculator, HarappaWorld) for people of 100% South Asian descent for example which does a decent job showing variation within the Indian subcontinent and its neighboring countries, but any substantial admixture from beyond Afghanistan or Bangladesh would disrupt it entirely.

So if they got an algorithm which could combine IBD/IBS analysis with this 'Oracle/Admixture' method, it would remove some of the inconsistency from their existing system. This is an entire degree more difficult however. The alternative is to use self-reported ancestry (i.e, where are your four grandparents from) along with a disclaimer that changing your grandparents' locations will change the model, so people can freely play around with it and thereby avoid some confusion when sharing results (like the "Confidence Level" setting in 23andMe). So people could be like "these are my results with all 4 grandparents set to England, here are my results with 2 grandparents set to Poland" and so on

vettor
07-23-2020, 05:58 PM
Its not just about some components appearing, but also the size of them. Their "ethnicity map" tool is helpful in checking this, since you can look at how many customers from a given country show a specific ancestral component. Actual Italian customers get the component "Italian" at 76,1 % (doesn't tell you how much of it, but seems rather low, even if considering non-Italians residing and buying from Italy) and "Iberian" at 41,1 %. "Balkan" and "Eastern European" is very, very close and can switch in individuals.

Italian % for my family

myself = 70.2

my father = 56.6 & 20.4 north and west European

my sons = 64 and 65 resp.. italian

my 1st cousin ( paternal side from my grandfather sister ) = 54.9 & 16 north and west European

my sister = 64


I think , like most admixture sites...they fail to produce a central european or a Alpine european percentage ....so these groups fall into different groups depending on site

Riverman
07-23-2020, 06:43 PM
Italian % for my family

myself = 70.2

my father = 56.6 & 20.4 north and west European

my sons = 64 and 65 resp.. italian

my 1st cousin ( paternal side from my grandfather sister ) = 54.9 & 16 north and west European

my sister = 64


I think , like most admixture sites...they fail to produce a central european or a Alpine european percentage ....so these groups fall into different groups depending on site

They fail with all more heterogenous populations, especially if they have no excellent references. That's a problem for all sites, but some produce even more inconsistent and unrealistic results.

Dorkymon
07-23-2020, 07:37 PM
I think I've mentioned this in other posts before, but MyHeritage definitely works something more similar to an oracle than anything else. It tries to find an optimal fit, rather than identify actual admixture, which can lead to people of the same ethnic background scoring in completely different categories. For example, some of my matches lose most of their 'Greek and Southern Italian' for a combination of Sardinian and Sephardic Jew instead. Obviously, these same people score 0% Sardinian on 23andme, because they aren't actually Sardinian. I've actually never experienced a good result on MyHeritage for any of the kits I've purchased for myself and family, which sits at 4 currently.

In my opinion, what matters most for the admixture test is if it's consistent between people of the same ethnic background. If everyone with similar ethnic background scores in the same categories, with similar levels of %, then I say the test is good. 23andme does this well. Even if I score mostly Italian instead of Greek, so do all of my matches of same ethnic background - the levels of Italian and WANA are very consistent for all Dodecanese islanders. MyHeritage on the other hand, has my matches in completely different categories in very different %. So I think its quality is poor.

I'm waiting on my AncestryDNA to rate that one, but it seems to have some issues also. We can only hope that MyHeritage's update is not just increasing the amount of regions, but actually fixing the consistency of their admixture results.

Yep, it's like an Oracle, but there are also components, which are too similar to one another. I score 89% Balkan, 3% East Euro, 6% West Euro and some 2% North Asian. There is a match of mine who scores like 60% Balkan and 40% East Euro, with some residue that I don't remember now, yet on Eurogenes K13 on GEDmatch we are identical.

Shadogowah
07-23-2020, 08:55 PM
I have seen people with real Jewish or Near Eastern ancestry and without on MyHeritage. Its being almost randomly assigned in the way described, oftentimes just as a minor admixture component to balance out a too North Western ancestral component for Central Europeans (like English, Irish, Scandinavian instead of general Northern and Western European). I could check it for cousins from the same clusters, one gets it, the other gets the more appropriate NW:Balkan:South Italian mix, which is just another admixture experiment and not identity too however. None of them has Jewish/Near Eastern on 23andme above trace levels and they are all fairly North Western in Oracle and on the PCA. So while its always possible the admixture is real, a hit on MyHeritage is the least reliable starting point.

If it is relevant for anything, I uploaded my FTDNA data to MH and later I bought a test. So they applied the algorithm twice to what should be the same markers.

In both cases the results gave more or less the same thing: 75% Iberian with a minor non-Spanish western mediterranean twist and two smaller strokes of MENA and Scandinavian+Finnish.

Main differences:

10% Italian was completely replaced by Sardinian that also took a 3% of the dominant Iberian.

My 7-8% MENA that was split in half between North African and Middle East now became something like 4 to 1 in favor of North Africa.

It also raised Scandinavian from 4% to 5.3%

As I guess it happens with most of people, I am mostly interested on those smaller percentages. I assume the Scandinavian admixture is some sort of ancestral stuff (no known Swedes, Britons, Irish or other northern Europeans in my genealogy tree) but due to the history of my country, I'd like to know more about the MENA admixture as it is not clear to me if it is also ancestral or I could actually have "real" Berber, Arab or Jewish medieval ancestry.

Riverman
07-23-2020, 11:50 PM
As I guess it happens with most of people, I am mostly interested on those smaller percentages. I assume the Scandinavian admixture is some sort of ancestral stuff (no known Swedes, Britons, Irish or other northern Europeans in my genealogy tree) but due to the history of my country, I'd like to know more about the MENA admixture as it is not clear to me if it is also ancestral or I could actually have "real" Berber, Arab or Jewish medieval ancestry.

You know, both your Scandinavian (Gothic, Frankish, French?) and MENA can be very real. But that's exactly the issue: If there is no consistency in segment assignment, but some components being made up to balance out for the synthetic admixture proportions, you can't bus sure about anything and have to try all kinds of other stuff to come to a meaningful interpretation.

Shadogowah
07-24-2020, 06:51 AM
You know, both your Scandinavian (Gothic, Frankish, French?) and MENA can be very real. But that's exactly the issue: If there is no consistency in segment assignment, but some components being made up to balance out for the synthetic admixture proportions, you can't bus sure about anything and have to try all kinds of other stuff to come to a meaningful interpretation.

I was thinking about the possibility that a combination of both Scandinavian+Finnish and MENA were just the result of that attempt to balance part of a southeuropean admixture that could not be made to fit in the existing models.

Riverman
07-24-2020, 09:01 AM
I was thinking about the possibility that a combination of both Scandinavian+Finnish and MENA were just the result of that attempt to balance part of a southeuropean admixture that could not be made to fit in the existing models.

Thats happening quite a lot with their estimates. And if you have to find out yourself whether its real or just a glitch, a synthetic constract or whatever you want to call it, its bad. I mean there are people from Bavaria which got no single percent from North Western Europe, but a lot of English and Scandinavian... plus a much too large Italian admixture. And their cousins got 90 percent North Western with basically the same ancestry. That's how it works, so you can never be sure what it is, if producing such absurd results.
Probably their algorithm is not THAT bad, but just their references are. But this kind of lack of recognising regional identities seems to be an error of the software too.

Kasana
09-04-2020, 02:49 PM
aheer of haryana results?

JFWinstone
09-04-2020, 03:41 PM
I currently only have results from uploads but here is a table of them. I've grouped together similar ethnicities, I think what is going on with my results especially with the East Asian is that the areas are overlapping so for example because our Chinese is South Chinese it is being picked up as Chinese & Vietnamese on one and Thai and Cambodian on the others. Their African regions still need some work, the lack of any Southern/Southeastern African groups is probably why a large part of it was assigned Nigerian. There isn't any that cover anything south of Congo and their "Central Africa" seems to cover quite a wide area. The Scandinavian still seems excessive.

https://i.imgur.com/54nJl9e.png