Automated Global25 nMonte entirely from the web

[poi]

Can any help give suggestions of ancient samples to try?

David said aim for between 1%-2% on this calculator but i can't seem to get much closer than this:

1.9572

Italy_Medieval_Collegno,50.83
Germany_Medieval,30.83
Hungary_Medieval_Szolad,15
Scythian_Hungary,2.5
Ethiopia_4500BP,0.83


Maybe the proper samples are just lacking for me?

Did you set penalty=0 ? I am seeing you a bit lower with no penalty

[Alstrohma]

Did you set penalty=0 ? I am seeing you a bit lower with no penalty

I haven't. I wasn't sure if it was appropriate. I will check.

[poi]

I haven't. I wasn't sure if it was appropriate. I will check.

If the goal is to lower fit distance, set the penalty to 0 and help nmonte go crazy with the minimization. Otherwise, it will penalize further reference pops. But having the penalty, the estimation might be better.

[Alstrohma]

If the goal is to lower fit distance, set the penalty to 0 and help nmonte go crazy with the minimization. Otherwise, it will penalize further reference pops. But having the penalty, the estimation might be better.

I'm basically looking for the closest but most realistic I can get

[poi]

I'm basically looking for the closest but most realistic I can get

1.9 is pretty good already then, assuming those references are historically reflective or close enough.

We need to categorize these references by region and epoch... Makes lives easier to sort through and try different possibilities.

[Alstrohma]

1.9 is pretty good already then, assuming those references are historically reflective or close enough.

We need to categorize these references by region and epoch... Makes lives easier to sort through and try different possibilities.

Its accurate to my knowledge. The only thing that is very strange is the Ethiopia. Seems like noise but could be real I guess.

Categorizing these would definitely make things easier. I'm not fully informed on all the proper populations to use and what is from what time era.

[JMcB]

This was kind of interesting. Although, I’m not sure it’s proper procedure. I ran some models using the two Roman Gladiators I’m supposed to be the closest to and then added the two I1a samples (my haplogroup) found among the Longobards.


England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.4524 — 19.17, 30.83, 12.5, 15, 10.83, 11.67


England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Germany_Roman +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.3184 — 20, 25, 15, 12.5, 5.83, 7.5, 14.7

And a more normal run:

England_Anglo-Saxon +England_Roman +Hungary_Medieval_Szolad +Italy_Medieval_Collegno +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.2026 — 4.17, 29.17, 25, 33.33, 8.33

[firemonkey]

This was kind of interesting. Although, I’m not sure it’s proper procedure. I ran some models using the two Roman Gladiators I’m supposed to be closest to and then added the two I1a samples (my haplogroup) found among the Longobards.


England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.4524 — 19.17, 30.83, 12.5, 15, 10.83, 11.67


England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Germany_Roman +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.3184 — 20, 25, 15, 12.5, 5.83, 7.5, 14.7

And a more normal run:

England_Anglo-Saxon +England_Roman +Hungary_Medieval_Szolad +Italy_Medieval_Collegno +Sweden_Viking_Age
Custom:AGUser_JMcB = 1.2026 — 4.17, 29.17, 25, 33.33, 8.33

1 England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age Custom:AGUser_firemonkey 1.663 60.83 0 14.17 0 17.5 7.5
2 England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age Custom:AGUser_firemonkey_father 2.2071 48.33 0 5.83 0 36.67 9.17


England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Germany_Roman +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age Custom:AGUser_firemonkey 1.6653 60.83 0 14.17 0 0 18.33 6.67
2 England_Anglo-Saxon +England_Roman--3DT16 +England_Roman--6DT18 +Germany_Roman +Hungary_Medieval_Szolad--SZ45 +Italy_Medieval_Collegno--CL63 +Sweden_Viking_Age Custom:AGUser_firemonkey_father 2.2032 49.17 0 4.17 0 0 36.67 10

1 England_Anglo-Saxon +England_Roman +Hungary_Medieval_Szolad +Italy_Medieval_Collegno +Sweden_Viking_Age Custom:AGUser_firemonkey 1.4332 20 14.17 25.83 35.83 4.17
2 England_Anglo-Saxon +England_Roman +Hungary_Medieval_Szolad +Italy_Medieval_Collegno +Sweden_Viking_Age Custom:AGUser_firemonkey_father 1.4606 0 20.83 4.17 72.5 2.5

[Capitalis]

I learn by doing and making mistakes (like a bull in a china shop). I read through the long G25 thread and found the explanation of why having the individuals in the data file rather than the averages allows nMonte3 to work its magic as this is built in to the design. So my use of averages was causing the Nordic_IA values to increase in people's results and I misunderstood what overfitting meant - I realise now that if you separate individuals from their parent group (eg. England_IA to England_IA_West, England_IA_East, etc.) this turns them into separate populations, rather than leaving them in their parent group and letting nMonte do the work.

I re-ran nMonte with the correct approach (my result is the unlabelled top bar).

Attachment 25913

Then the Anthrogenica members:

Attachment 25912

Whew.

[poi]

I learn by doing and making mistakes (like a bull in a china shop). I read through the long G25 thread and found the explanation of why having the individuals in the data file rather than the averages allows nMonte3 to work its magic as this is built in to the design. So my use of averages was causing the Nordic_IA values to increase in people's results and I misunderstood what overfitting meant - I realise now that if you separate individuals from their parent group (eg. England_IA to England_IA_West, England_IA_East, etc.) this turns them into separate populations, rather than leaving them in their parent group and letting nMonte do the work.

I re-ran nMonte with the correct approach (my result is the unlabelled top bar).

Attachment 25913

Then the Anthrogenica members:

Attachment 25912

Whew.

nMonte3 is awesome in actually utilizing individuals and the reporting is at the group level anyway. That is the primary reason why the tool did not even bother with calculating averages during the Step1-Reference Groups. Feed in all individuals and let nmonte3 sort it out. It might also be interesting to check against the average, which you can do in the Test step2 though.

Btw, your charts make it seem all academic