PDA

View Full Version : Gedmatch - "generations"



Dante
04-10-2016, 11:02 PM
Just curious what is the generational breakdown of the populations for gedmatch.

I get the smaller the number = closer DNA relation you are.

But is there any calculator and/or way of figuring the X generations that separate you.

For example

Kent @ 2.23 - could I take that as there's 2.23 generations between myself and the general public sampled from that region?


I am looking at it sort of akin to the cousins / relative breakdown you can get elsewhere where it'll tell you X generations, potential, between you and so and so.

AJL
04-11-2016, 12:04 AM
Just curious what is the generational breakdown of the populations for gedmatch.

I get the smaller the number = closer DNA relation you are.

But is there any calculator and/or way of figuring the X generations that separate you.

For example

Kent @ 2.23 - could I take that as there's 2.23 generations between myself and the general public sampled from that region?


I am looking at it sort of akin to the cousins / relative breakdown you can get elsewhere where it'll tell you X generations, potential, between you and so and so.

No no no, it's not a generational calculator, it's a measure of distance based on squared differences of mutations. "1.00" would mean there's a perfect correlation between you and the population set -- and I've never seen this happen with anyone. 2.23 is a very good match for a single population. As an example, for puntDNAL K12modern, the lowest I get for one population is 6.17, and with Dodecad V3, it's 11.0 for one population.

Dante
04-11-2016, 01:17 AM
Oh okay. Just it seemed really bizarre that it got a few populations right on the "dot" for grandparents, etc.

In the oracle
Mixed Mode Population Sharing:
What exactly would this mean??
1 79.5% Danish + 20.5% Spanish_Valencia @ 1.21


And I know autosomal / recombination play a factor but is there a reason why full siblings would have quite a difference in oracle results.

I mean EU test my 1st is southeast_English, but 8th for sibling.

AJL
04-11-2016, 02:29 AM
That means there's very high confidence that your mixture places as 8/10s Danish and 2/10ths Valencian. Using your weighted values on a map, it places you on average around NW Germany. That's not inconsistent with a sister who places in Kent, believe it or not.

The more diverse one's ancestry, the more different two siblings can be. My aunt happened to inherit pretty well all my ancestors' Scottish and Irish genes, so she shows up around Down/Argyll. My mother picked up proportionally more of their Welsh, French, German and Dutch ancestry as well as some trace of something that doesn't seem to belong in Europe, so she places around the English Channel or north Belgium. They almost never place in the same country with any analysis, despite showing as full siblings at 23andme.

vettor
04-11-2016, 08:25 AM
No no no, it's not a generational calculator, it's a measure of distance based on squared differences of mutations. "1.00" would mean there's a perfect correlation between you and the population set -- and I've never seen this happen with anyone. 2.23 is a very good match for a single population. As an example, for puntDNAL K12modern, the lowest I get for one population is 6.17, and with Dodecad V3, it's 11.0 for one population.

it depends which you test with

with MDLP K13 Ultimate I get
# Population (source) Distance
1 Italy-Friul 2.94
2 Swiss 3.79
3 Occitan 4.16
4 Italy_North 4.82
5 Italian_Piedmont 4.84
6 Italian_Bergamo 5.61

with Dodecad V3 I get Tuscan at 8.1

Where does the "good match" end ?..........at 3 or 4

Dante
04-11-2016, 11:01 AM
That means there's very high confidence that your mixture places as 8/10s Danish and 2/10ths Valencian. Using your weighted values on a map, it places you on average around NW Germany. That's not inconsistent with a sister who places in Kent, believe it or not.

Thanks. This helps. I've played with gedmatch before but never really looked at it. Genetics isn't my thing. Probably cause English isn't my first lanuage.


Okay that makes sense for location.

Got one more question.

In the oracle 4 it gives you
2 populations
3 populations
4 populations

I figured from what I've found 2nd populations would be your "parents" and 4 populations "grandparents". However, what is the 3 populations?


Similarly, derived would be what with some populations? Is that for samples taken from actual locations [e.g. swedish north (derived) - would be from actual northern sweden]??

Also there's one oracle which has both English & British are populations. That threw me for a loop. The difference would be what exactly? English from England? British from the entire UK? Or the other way around?

AJL
04-11-2016, 01:02 PM
Where does the "good match" end ?..........at 3 or 4

It seems to be a little different relative to each calculator. Tuscan is farthest away from you of all those, but OTOH Dodecad 3 just seems to have lower confidences too.

AJL
04-11-2016, 01:04 PM
In the oracle 4 it gives you
2 populations
3 populations
4 populations

I figured from what I've found 2nd populations would be your "parents" and 4 populations "grandparents". However, what is the 3 populations?

This would be a parent of one population, and the other parents who is a mix of two equal populations. (So one parent's spot weighted at 50%, the other at 25% each for each of their ancestries.)



Similarly, derived would be what with some populations? Is that for samples taken from actual locations [e.g. swedish north (derived) - would be from actual northern sweden]??

Also there's one oracle which has both English & British are populations. That threw me for a loop. The difference would be what exactly? English from England? British from the entire UK? Or the other way around?

I'm not sure what "derived" referto in this context.

"English" is more precise -- England, while "British" can include Scots and Welsh.

Dante
04-12-2016, 11:19 AM
I'm not sure what "derived" referto in this context.

MDLP World-22 - oracle 1 / mixed populations - has every population followed by (derived) in brackets like that.

AJL
04-12-2016, 03:38 PM
Thanks. He explains on his site:

"Please notice: 'ancestral' populations (i.e 'simulated populations' from the previous step - see above) are labeled in Oracle results as (anc), while the 'real world' modern and ancient populations are marked as 'derived.'"

Dante
04-13-2016, 10:28 AM
Thanks you've been quite helpful. Like said English isn't the first language & I've played with gedmatch just never really "dove in" till now.