Ancestry Sideview Ethnicity Results - How Accurate?

[CliveR]

I am interested to find out how accurate people think Ancestry's Sideview is, particularly with regard to ethnicity estimates. Is it to be taken with a pinch of salt?

While I am no DNA expert, I find Sideview's separation of paternal and maternal ethnicities rather hard to believe from a common-sense point-of-view.

Let me first say that in researching my family tree, I have found that my mother's ancestors and father's ancestors come from roughly the same region of the UK, mostly in the North West of England:

Paternal ancestors: Cheshire, South Lancashire and Shropshire
Maternal ancestors: Cheshire, South Lancashire and Northern Ireland

I'd say that my maternal family is more biased towards South Lancashire than my paternal family, which is more Cheshire-oriented.

My overall ethnicity results appear to be very typically English, but it is their separation to paternal and maternal by Sideview that concerns me.

Scotland
In my overall ethnicity results, I was surprised to see 18% Scotland but I believe this is not untypical for an English person. Ancestry say "on average, people native to England who take a DNA test see about 20% Scotland in their results". However, Sideview attributes all of my Scottish to my father's side. Surely, if the Scottish came from ancestors in the dim and distant past, it should be roughly equal on both sides, especially since my parents come from the same region of England.

Scandinavia
It's a similar story with my 6% Norway. Again, I understand this to be typical for an English person and have heard it described as "Viking residue". However, Sideview has attributed it purely to my mother's side. If it was Viking residue, I would expect it to be on both sides - for clarification, there are no obvious Norwegian ancestors in my family. I do, however, have Sweden & Denmark on both sides, 1% from my mother and 3% from my father.

Ireland
Another puzzle is my 4% Ireland. I know that I have Irish ancestors on my maternal side but Sideview has attributed it to my paternal side, with none on my maternal side. Could I be completely missing 6.25% of Irish ethnicity from a maternal great-great grandmother?

My main concern is that Scotland and Norway are polarised in my Sideview results, and I don't believe they should be if they come from the very distant past. Has anyone else observed similar effects in their results?

 

[peter]

Forget ethnicity estimates - they are a waste of time, whether split by parent or not.

Scandinavia
It's a similar story with my 6% Norway. Again, I understand this to be typical for an English person and have heard it described as "Viking residue".

It's a rather misleading term since most people will associate Vikings with the raids, rather than the Norman Conquest. Best ignored.

 

[CliveR]

Forget ethnicity estimates - they are a waste of time, whether split by parent or not.

I know that it is not an exact science but I would like to know whether other people have noted similar results with the Sideview feature.

 

[Katie Bee]

I have Northern Irish ancestry on my paternal side and that is reflected in Ancestry Sideview Ireland 12% and Scotland 41%
Ancestry's Scotland covers Northern Ireland, and my ancestors were Protestants, so I was assuming they came from Scotland originally, although I have no documentary evidence up to now.
My maternal side says 29% Scotland and again I have no documentary evidence for that either.
I have 2% Sweden and Denmark on my paternal side, which if I had to guess I would have put on my maternal side. So I was not expecting that.
England and NothWestern Europe only 4% on paternal side and 38% on maternal side. I would have expected that to be more even.

 

[peter]

I know that it is not an exact science but I would like to know whether other people have noted similar results with the Sideview feature.

There will be people who have found the results plausible and there will be others who have found them implausible. And then when the estimates change, with the next iteration of Ancestry's algorithms, it could be the other way round.

Ethnicity estimates were a long way from being useful even before the SideView split - now they are even less useful, because they provide more opportunities to waste our tome and effort. On many occasions I have said that they are for amusement only, and that remains the case for anyone who has researched their family tree, especially to the sort of depth that most LostCousins members have.

I'll be delighted when ethnicity estimates have got to the point where they are useful, but the fact that there has been no consistent or continuous improvement over the past few years suggests that there's a long way to go, at least for those of mainly British and Irish heritage.

 

[uncle024]

DNA never lies however ethnicity estimates are just estimates based on all the other people who have taken a test. You don’t say if you have matches to known paternal & maternal cousins and if so, they have been placed on the correct side of the family. With adoption you have to tell the system whether Parent1 or Parent2 is Maternal. I say Maternal because generally it’s the mother who is known via original birth certificate.

 

[CliveR]

I have Northern Irish ancestry on my paternal side and that is reflected in Ancestry Sideview Ireland 12% and Scotland 41%
Ancestry's Scotland covers Northern Ireland, and my ancestors were Protestants, so I was assuming they came from Scotland originally, although I have no documentary evidence up to now.

Thanks for your response, Katie.

I hadn't noticed that Ancestry's Scotland covers a part of Northern Ireland, but now I see it. I believe that my maternal Northern Irish ancestors were Catholic, so does that mean they probably didn't migrate from Scotland? In any case, Sideview has the Scottish and Irish on my paternal side, on which I know of no such ancestors.

 

[CliveR]

DNA never lies however ethnicity estimates are just estimates based on all the other people who have taken a test. You don’t say if you have matches to known paternal & maternal cousins and if so, they have been placed on the correct side of the family. With adoption you have to tell the system whether Pareten1 or Paretn2 is Maternal. I say Maternal because generally it’s the mother who is known via original birth certificate.

Yes, I do have DNA cousin matches to known sides and Sideview has largely assigned them to the correct side, although I currently have a large Unassigned group, like most people (in some cases, the side they belong to is very obvious to me). There is a branch of my tree where some of my paternal cousins have been assigned to the maternal side and I can't see any obvious way that there has been cross-over between the two sides. However, Sideview is supposed to be at least 95% correct for 90% of people, so the anomalies may be within the margin of error.

By the way, I wasn't disputing that ethnicity estimates are estimates. I was questioning whether Sideview actually gives sensible results when applied to ethnicity and wanted to know whether other people were seeing the same polarisation effects as me for supposed distant ancestors.

A related issue is the Chromosome Painter. Here, I have mostly chromosomes of solid ethnicity, including whole chromosomes of Scotland, Norway and Sweden & Denmark. Ancestry say:

"With this data, you can get a general feel for how recent your connections are to your ethnicity regions. The longer a segment (a single colored block) is for a region, the more recent your connection to that region may be."

So, I would expect to see fragmented chromosomes containing multiple ethnicities from my distant Scottish and Viking ancestors. At most, I have two ethnicities in one chromosome.

 

[peter]

So, I would expect to see fragmented chromosomes containing multiple ethnicities from my distant Scottish and Viking ancestors. At most, I have two ethnicities in one chromosome.

Don't expect sensible results - it's a bit of fun.

 

[Stuart]

The word "expect" is used a lot in this thread, about what Ancestry calls "ethnicity". I reckon most people are basing what they expect on very optimistic assumptions, and forgetting how much randomness is involved in all our DNA. Even if whatever Ancestry's processing responds to is something like the historical/geographic identities that the ethnicity labels suggest (and I doubt it is), an individual's DNA will not in most cases match that closely. And here is an illustration of this.

A few weeks ago Ancestry sent out some e-mails pushing their new "DNA compare" feature. (There is a "Compare my DNA" button in the Ethnicity estimate box.) One thing you can do with this is a direct side-by-side comparison of the ethnicity figures for two siblings among your DNA matches. Of course you may need to do a bit of work to find out who are siblings, since they won't often say so explicitly. They and their parents will usually be living now or recently, and only limited details may be given. But with luck you'll find a few examples. Here's one from my matches.

These two brothers have a Polish father and a mother who was herself half English and at least a quarter Scottish in terms of recent family. Since she was from Durham, the Scottish component of her ancestry could be much higher. That explains the major features of these two sets of figures. But they don't correspond exactly to the "expected" figures of 50%, 12% etc. And I don't think you should expect them to either. These are two random fluctuations around a trend, and that's all you'll ever get. Of course your pair of siblings may match closer than these, or be even more different. But I expect this is typical.

 

[uncle024]

These are two random fluctuations around a trend, and that's all you'll ever get. Of course your pair of siblings may match closer than these, or be even more different. But I expect this is typical.
View attachment 2339

Good points Stuart. The thought occurs to me that in the above example the actual amount of 'Scotland' DNA in cM may actually be the same. But the percentage gets skewed by the right hand sibling having more English, Sweden and Eastern European. In the right hand example 12% less Scotland but 9% + 8% =17% more English & Sweden.

 

[peter]

The thought occurs to me that in the above example the actual amount of 'Scotland' DNA in cM may actually be the same. But the percentage gets skewed by the right hand sibling having more English, Sweden and Eastern European. In the right hand example 12% less Scotland but 9% + 8% =17% more English & Sweden.

That's not how it works - if the amount of DNA is the same then the percentage will be the same.

There are similar differences between my brother's ethnicity estimates and my own, and the difference is even more stark when you look at three siblings who are cousins of ours. The first two columns are me and my brother - note that he has 13% from Sweden and Denmark compared to my 0%.

The last three columns are our cousins (2nd cousins once removed, as it happens, but that's not relevant to this discussion). Note how their Scotland figure varies from 2% to 23%.

Seeing the three sets of figures side by side you can rationalise why there are differences, but imagine that you only saw the figures in column 4 - how representative would it be?

 

[Helen7]

A set of 3 siblings amongst my matches show differences between them, but they are less stark than those of Peter's cousins.

 

[Katie Bee]

I believe that my maternal Northern Irish ancestors were Catholic, so does that mean they probably didn't migrate from Scotland?

Sorry CliveR I do not know the answer. I understood that they were mainly protestant faiths that moved to Ulster from the 1600s onwards.

 

[Katie Bee]

These are the ethnicity results for me and my sister.
The only strange thing is that all the Irish for me is on my paternal side and all the Irish for my sister is on her maternal side.
As Peter says

Forget ethnicity estimates - they are a waste of time, whether split by parent or not.

 

[CliveR]

That's all very interesting and helps to demonstrate the randomness of inherited DNA and perhaps the flakiness of ethnicity estimates. My sister's DNA sample is currently being processed, so it will be interesting to see how hers compares with mine.

Ancestry also have 'DNA communities'. Although I know that I have great-great-grandparents from Shropshire on one side and from Northern Ireland on the other side, neither of these regions feature in my communities. They don't feature for my first cousin and relevant second cousins from the respective sides either.

The above discussion relates to ethnicity estimates in general and was a bit of a deviation from my initial question about Sideview. If, for a moment, we accept our overall ethnicity estimate as it is, do we think Sideview does a reasonable job, particularly with respect to distant ancestors such as the Vikings? It would be interesting to see some screenshots of people's Sideview breakdowns. I have only seen my own, as Ancestry doesn't give access to other people's Sideview results.

 

[peter]

The above discussion relates to ethnicity estimates in general and was a bit of a deviation from my initial question about Sideview. If, for a moment, we accept our overall ethnicity estimate as it is, do we think Sideview does a reasonable job, particularly with respect to distant ancestors such as the Vikings? It would be interesting to see some screenshots of people's Sideview breakdowns. I have only seen my own, as Ancestry doesn't give access to other people's Sideview results.

Since overall ethnicity estimates are of little or no value (or even negative value if they distract you from what really matters), it follows that the SideView breakdown is even less useful.

The only reason I am posting in this discussion is to ensure that nobody who might come across it now or in the future will be under any illusions!

 

[CliveR]

Since overall ethnicity estimates are of little or no value (or even negative value if they distract you from what really matters), it follows that the SideView breakdown is even less useful.

The only reason I am posting in this discussion is to ensure that nobody who might come across it now or in the future will be under any illusions!

I am not advocating Sideview. The nature of my post is to question how successfully it achieves its aim (at this Beta stage) in relation to paternal/maternal ethnicity breakdown. My view is it probably doesn't do it very successfully, but I was hoping to learn about other people's results and views on the matter.

 

[peter]

I am not advocating Sideview.

Perhaps you should - it is very useful when it come to matches with cousins, provided you regard it as a strong hint, rather than evidence. But it isn't any use for ethnicity because no matter how good the SideView algorithm is, when it's starting from a set of data that is essentially useless it's likely to make matters worse.

The nature of my post is to question how successfully it achieves its aim (at this Beta stage) in relation to paternal/maternal ethnicity breakdown. My view is it probably doesn't do it very successfully, but I was hoping to learn about other people's results and views on the matter.

You've already had the same answer from several people. Bearing in mind the randomness of DNA inheritance and the erratic nature of ethnicity estimates, you may well find someone eventually who has formed a different impression, but that isn't a good reason to prolong the discussion.

The time to revisit the question will be when Ancestry next revise our ethnicity estimates, but bear in mind that most of their customers are in the New World, some of them with very different genetic make-up (you only have to look at the examples they give on their website, and in YouTube videos).

 

[CliveR]

You've already had the same answer from several people. Bearing in mind the randomness of DNA inheritance and the erratic nature of ethnicity estimates, you may well find someone eventually who has formed a different impression, but that isn't a good reason to prolong the discussion.

The time to revisit the question will be when Ancestry next revise our ethnicity estimates, but bear in mind that most of their customers are in the New World, some of them with very different genetic make-up (you only have to look at the examples they give on their website, and in YouTube videos).

OK, it sounds like you've decided that this is the end of my discussion (for now).