PDA

View Full Version : How results across tests differ: which to trust most?



memianec88
03-27-2018, 12:30 PM
Hello! So far, I have run my raw data through several services: FTDNA, Gedmatch, DNA.Land, and My Heritage. They pick up on different things, which is interesting, but only Gedmatch picks up on West Asian and South West Asian DNA (at approximately 10% and 3% respectively) and does so consistently across calculators. What should I think of this? Thank you for your insights!

Kale
03-27-2018, 05:27 PM
The calculators at Gedmatch seem to suffer from the calculator effect, see here...
http://bga101.blogspot.com/2012/05/beware-calculator-effect.html
and
http://dodecad.blogspot.com/2012/08/on-so-called-calculator-effect.html

And should probably be disregarded for any fine scale ancestry.

A Norfolk L-M20
03-27-2018, 09:12 PM
None of them, and all of them. Welcome to ethno-geographic testing. GEDmatch isn't a test. It is a platform that enables developers to float different calculators. These calculators can become addictive, as developers gather more interesting references, and measure genetic distances using different systems.

Discovering ancestry, by using DNA testing - often in combination with other methods such as family lore, recorded genealogy, or local history - can be addictive. Stay around an embrace the learning curve.

RobinBMc
03-27-2018, 11:01 PM
Do you have known West or South West Asian ancestry?

memianec88
03-28-2018, 02:09 AM
The calculators at Gedmatch seem to suffer from the calculator effect, see here...
http://bga101.blogspot.com/2012/05/beware-calculator-effect.html
and
http://dodecad.blogspot.com/2012/08/on-so-called-calculator-effect.html

And should probably be disregarded for any fine scale ancestry.

Thank you for the articles! Pardon me, What is fine scale ancestry? :):)

memianec88
03-28-2018, 02:10 AM
None of them, and all of them. Welcome to ethno-geographic testing. GEDmatch isn't a test. It is a platform that enables developers to float different calculators. These calculators can become addictive, as developers gather more interesting references, and measure genetic distances using different systems.

Discovering ancestry, by using DNA testing - often in combination with other methods such as family lore, recorded genealogy, or local history - can be addictive. Stay around an embrace the learning curve.

Addictive, indeed! Any tips on finding the lore part???

memianec88
03-28-2018, 02:13 AM
Do you have known West or South West Asian ancestry?

No, I’m a beginner. So far my paper trail includes Germany, Holland, Prussia, Poland, and Poland-Germany/Poland-Russia. I have some ancestors whose papers I haven’t chased down yet :)

RobinBMc
03-28-2018, 03:24 PM
No, I’m a beginner. So far my paper trail includes Germany, Holland, Prussia, Poland, and Poland-Germany/Poland-Russia. I have some ancestors whose papers I haven’t chased down yet :)

Then I wouldn't put much stock in the West/South West Asian results, at least not in terms of recent genealogy. Plus, depending what calculators you were using, the results might have been wrongly skewed towards those regions and not an accurate representation of your genome.

Kale
03-28-2018, 03:35 PM
Thank you for the articles! Pardon me, What is fine scale ancestry? :):)

Trying to discover very small contributions to your ancestry (like if you were 2% Native American) or trying to discern between very similar ethnicities (like trying to tell whether grandma is English or Scottish)

memianec88
03-31-2018, 12:01 AM
Aha! Thank you for the explanation :)

memianec88
03-31-2018, 12:02 AM
Trying to discover very small contributions to your ancestry (like if you were 2% Native American) or trying to discern between very similar ethnicities (like trying to tell whether grandma is English or Scottish)

Oh, I almost forgot to ask, would you consider ~10% fine scale?

Saetro
03-31-2018, 02:00 AM
Addictive, indeed! Any tips on finding the lore part???

I was lucky. My grandmother and her sister would tell me stories about family when I was young, so I went to them first.
Basically you need to talk to anyone you can.
Hunt down tips on Oral History.
Sorry, the websites I used to get started a decade ago have moved on.

Go and talk to the elders.
Allow for lots of extra time if you hit it off and they want to extend the visit - don't have something you need to rush off and do later.
Ask open questions.
Take things like photos.
(Heirlooms can be good, but can backfire if something that went to your line instead of theirs has been resented.)
Ask them to describe their house as it was when they came home from school and walked into and through it.
Or if you want to know about the neighbourhood, about the journey from home to school.
(My mother thought of that one and took us on a wonderful walk through her childhood on the way.)
Use a mobile phone or computer to record audio.
(If you can obtain a lapel mike for the subject it can capture their responses wonderfully, but it might not hear your questions - so write them down beforehand or as you go.
I bought a recording setup intended for capturing live music gigs that had two inputs for two lapel mikes - ideal. Some Oral History societies used to rent similar gear out to members.
This gear may seem excessive but the quality is awesome and segments go well in a media presentation later.)
A serious Oral Historian does a number of extra things.
One I found was brilliant was to return a transcript copy to the interviewee, asking for permission to use their answers in whatever project I was working on.
Never had permission refused, but more than once amendments or additional information surfaced that was absolute gold dust.

Golden rules.
NEVER interrupt someone giving you an answer if at all possible. Take down a note to ask for clarification later.
NEVER contradict.
NEVER tell someone that was not how it happened. Your other version may be wrong. Or two different people may have differing perceptions (and neither may be right!).
The best you can do is to mention that you thought you remembered a different version, but that you had probably misheard/misremembered/misunderstood.
Sometimes people will not be offended and explain why there might be a difference.

It may take an introductory approach first - from another family member or a first visit.
Your interviewee will be wondering why you have turned up - often out of nowhere - and what is in it for them.
All humans have a fundamental need to tell their story. That is usually enough - to be listened to.

You seem a thoughtful and considerate person.
I'm sure you will be fine.
(Final thought - take a cake.)

Kale
03-31-2018, 02:25 AM
Oh, I almost forgot to ask, would you consider ~10% fine scale?

Depends how different that 10% is you are suspecting. If you suspected you were 90% English and 10% Japanese, it would be pretty easy to find out you have 10% something East Asian. If you suspected you were 90% English and 10% Scottish, that would be harder because English and Scottish are similar relatively speaking.

Messier 67
04-05-2018, 01:20 PM
How do they determine whether a gene for instance German or not. Do they take all the German samples they have and if the gene is very high compared to the rest of the samples, then they call that a German gene.

Would it not be better to allocate points, this gene if you have it... will get you 10 German points, 3 Scandinavian points, 2 English points and 1 point for Netherlands. Based on statistics in the population. Then when tallying, if some locations have less than 5% of total, don't count them, or put them in the trace section. The places where you have 5% or greater will end up being individually 1% or more of your total. If some place has 50% of the points, they would get about 55% of the displayed results.

Is this system better than a winner takes all. Like in an election, after the election (DNA testing), everyone is informed they voted for the winner of the election, when this is not true for everybody.

RobinBMc
04-05-2018, 03:25 PM
How do they determine whether a gene for instance German or not. Do they take all the German samples they have and if the gene is very high compared to the rest of the samples, then they call that a German gene.

Would it not be better to allocate points, this gene if you have it... will get you 10 German points, 3 Scandinavian points, 2 English points and 1 point for Netherlands. Based on statistics in the population. Then when tallying, if some locations have less than 5% of total, don't count them, or put them in the trace section. The places where you have 5% or greater will end up being individually 1% or more of your total. If some place has 50% of the points, they would get about 55% of the displayed results.

Is this system better than a winner takes all. Like in an election, after the election (DNA testing), everyone is informed they voted for the winner of the election, when this is not true for everybody.

It's not that simple. Firstly, they don't look at individual genes - they look at combinations of SNPs. Secondly, I know at AncestryDNA they actually run 40 different analyses and then average out the results, and then provide a percentage range showing the lowest and highest results you got during the 40 different analyses.