DNA Ethnicity Nonsense

[Bob Spiers]

So given these estimates based on research of the time, here is my latest Ancestry Ethnicity update

England (& Northwestern Europe) 57% :: Scotland 17% :: Ireland 11% :: Wales 9%:: Norway 6%

Still with the same sceptical eye, and following on from my own ethnicity estimates (depicted in bold above) - I have now checked out my daughter's DNA ethnicity. Unlike me, she was not born in the Midlands and although that will naturally follow through as part of her ethnic route, but born as she was to my first wife I would expect to find her mother's strong ties to Bucks/Beds/Northants showing, but that does not happen. Here is how Ancestry shows her ethnic breakdown.

82% England & NW Europe. (Greater London, The Midlands, West Midlands) :: 11% Wales :: 3% Scotland :: 2% Ireland :: 2% Sweden

"Greater London" is a pretty poor "catch-all" and fairly meaningless if meant to cover North Buckinghamshire , West Bedfordshire and South Northamptonshire which predominate in my first wife's Tree. I can say this with some emphasis after researching her line as - almost boringly - one is hard pressed to find any connection -paternal or maternal - outside of this region, other than a few 'minnows'. I can accept the southern areas of Bucks come with the Greater London reach, but the others merely emphasise that Ancestry's attempt to delineate specific areas should be taken with a great deal of salt.

For the remaining comparisons I would accept Scotland and Ireland percentages to reduce as I found none other than my own within my daughters Tree, but cannot quite understand why her Welsh contingent has increased 2% from my 9% to her 11%. Perhaps that comes under the 'minnows' element and I will check this out further.

 

[peter]

A reminder that I have on many occasions described ethnicity estimates as being for amusement only!

 

[Helen7]

I'm not an expert on Irish history, but I would have thought religion is best pointer - I suspect that most protestants in Ireland had Scottish ancestry. Also, whilst I don't have any Irish ancestry, I do have at least one relative in my tree who was born in Ireland (to an English father who was serving in the army). So that's something else to consider.

Yes, that's a good point. He was married in an Anglican Church (in 1873) but other than that I don't know anything about his background.

 

[Alexander Bisset]

One thing that most people don't notice is the little arrow to the right of the percentages. Click that and you get the REAL numbers the RANGES the percentage cover. For most of the cases with <5% it will be 0-6% or something like that. For the larger numbers the percentages can be WILDLY variant. eg: a 17% Scotland could be 0-30%, which from a scientific point of view is nonsense.

Whilst the scientific method is sound they are simply working with too many variables and not nearly enough reference data (known test samples from known geographic populations from specific documented studies). The result is wide ranges that are more of a guess than an estimate. So basically the result is a marketing ploy.

I say marketing ploy as the very simple easy to understand percentages makes for an easy sales pitch, and that undoubtedly drives sales. Yet at the same time it is extremely deceiving as a ploy because these guesses are not robust enough to stand up to scrutiny, and give the casual observer totally false impressions. These are things like the false notion these numbers show your DNA changes, the false notion that it shows the tests can't be accurate because twins get different "results", etc. (Actually they are so imprecise that you can do the same test twice with Ancestry at the same time and get different ethnicity estimates).

The other major issue is that people think these meaningless numbers are their results. I've lost count of the number of posts I've seen online along the line of "OMG my DNA results have changed what does it mean". Sadly because of the marketing many people don't actually realise that the real results are the raw numbers and the matches you get from them. I'd be shocked if any LostCousins forum readers were that ill informed so it may be a revelation to readers that many Ancestry DNA testers don't know there is a matches section to their results. Many of these same people don't realise the envelope top right with a number on it shows they have waiting messages. This may go a long way to explain why so many matches we contact have no tree and whose messages go unanswered.

I'd love to start a movement to get the word out so we can re-educate people about shifting their focus to matches, including uploading a skeleton DNA tree and answering messages, and ignoring the random ethnicity estimates that can and do change rapidly over time. That change of focus for the large percentage who seem to think the estimates are their results could be transformative.

 

[peter]

One thing that most people don't notice is the little arrow to the right of the percentages. Click that and you get the REAL numbers the RANGES the percentage cover. For most of the cases with <5% it will be 0-6% or something like that. For the larger numbers the percentages can be WILDLY variant. eg: a 17% Scotland could be 0-30%, which from a scientific point of view is nonsense.

Whilst the scientific method is sound they are simply working with too many variables and not nearly enough reference data (known test samples from known geographic populations from specific documented studies). The result is wide ranges that are more of a guess than an estimate. So basically the result is a marketing ploy.

What's not clear is whether the range quoted indicates solely the statistical error, or also takes into account the randomness in our inheritance of DNA. For example, if two of my grandparents were 100% Scottish and two were 100% English the amount of Scottish DNA I inherit would be precisely 50% if the two Scots were married to each other, but anywhere between 0% and 100% if they weren't.

 

[Alexander Bisset]

What's not clear is whether the range quoted indicates solely the statistical error, or also takes into account the randomness in our inheritance of DNA. For example, if two of my grandparents were 100% Scottish and two were 100% English the amount of Scottish DNA I inherit would be precisely 50% if the two Scots were married to each other, but anywhere between 0% and 100% if they weren't.

Indeed and if they were matching in the same way as individual DNA matches that would be true. However it's a binary search they are looking for the presence or absence of certain groups of genetic markers, they are not looking for shared segments or similar sort of thing.

Thus the presence of one of the many markers they are looking for would indicate a Scottish connection. There will be quite a bit of randomness in there but probably not as stark as you suggest as it's not the same as checking for cM shared.

 

[SallyB]

I have been left rather perplexed by the recent ethnicity update as I am now 35% Scottish. On paper, in all the lines of my tree that I have been able to research, none of my ancestors even so much as went on a day trip to Scotland and all come from a fairly narrow geographical area: Nottinghamshire, Lincolnshire, Leicestershire, South Yorkshire and the East Riding. I realise that the ethnicity estimates are just that and not an exact science, but can anybody advise me where to go from here? I'm wondering if somebody in my tree isn't who they were supposed to be, or should I just ignore this latest update altogether?

 

[peter]

I have been left rather perplexed by the recent ethnicity update as I am now 35% Scottish. On paper, in all the lines of my tree that I have been able to research, none of my ancestors even so much as went on a day trip to Scotland and all come from a fairly narrow geographical area: Nottinghamshire, Lincolnshire, Leicestershire, South Yorkshire and the East Riding. I realise that the ethnicity estimates are just that and not an exact science, but can anybody advise me where to go from here? I'm wondering if somebody in my tree isn't who they were supposed to be, or should I just ignore this latest update altogether?

Ignore the latest figures and all the ones that came before - they are "for amusement only" as I've said in the newsletter on many occasions.

Follow the advice in my Masterclass and focus on your DNA matches instead. If you really do have Scottish ancestry it will soon become apparent.

 

[JimP]

I also had a significant percentage, 29% (0-29 range) of Scottish show up in my updated analysis, up from a much smaller percentage. I do have some Scots ancestry -- one (or possibly 2) POW from the Battle of Worcester and several Great Migration ancestors from Yorkshire and Lancashire on my father's side, a lot of Scots-Irish on my mother's side. The "predicted" amount would, however, be on the low end of the range.

For my husband (a Newfoundlander), his update also showed 33% Scottish (1-33 range). For a Newfoundlander from the west coast of the island, the gene pool is very limited: predominantly southwestern England (Somerset, Dorset, Devon, and a little bit Gloucester), with a little bit of French and Miq'maq. Predicted Scottish ancestry from his tree is nil.

As Peter says, "for amusement only". But it does seem that Ancestry's "Scotland" reference data is perhaps flawed.

 

[DavidF]

I watch a number of 'Family History' TV shows - such as WDYTYA. They often bring on an Ancestry expert who only talks about the ethnicity results - never the actual DNA matches! Several times I've seen (in the Australian version) the expert say that the ethnicity indicates Devon or Cornwall and then they immediately suggest that is where the person needs to go to find answers. I'm not sure which planet these people are on.
Also there was a case recently that was investigating a woman born in a German concentration camp. She knew her mother was Jewish but had no knowledge of her father. The ethnicity showed roughly 50/50 Jewish and German, so the expert stated that as she knew her mother was Jewish her father must be German. Later in the show they even named the supposed father based on DNA evidence, even though this person had no other descendants to match with.

 

[peter]

I watch a number of 'Family History' TV shows - such as WDYTYA. They often bring on an Ancestry expert who only talks about the ethnicity results - never the actual DNA matches! Several times I've seen (in the Australian version) the expert say that the ethnicity indicates Devon or Cornwall and then they immediately suggest that is where the person needs to go to find answers. I'm not sure which planet these people are on.

It's television.