Viking/Celt Model Mainly for Northwest Euros but other can try it (Scaled)

[Grace O'Malley]

It is just as bad as a underfitted model (with too high dist), the difference is that a overfitted model extracts noise from the data as if it represents real variation, while underfitted models fails to roughly corresponds to the data distribution. In a nMonte context the idea is to not use too many closely related populations, and in your model you are using 90 very closely related references, which makes it total crap in terms of statistical power.

I do understand that but being Irish these samples are realistic for my ancestry. A lot of the people that came were quite similar which explains why there hasn't been enormous changes since the Bronze Age but thanks for taking the time to explain this.

[Grace O'Malley]

Target: Kevin_scaled
Distance: 1.3917% / 0.01391715
30.4 Swede
28.4 Celt
18.0 Gallo-Roman
14.2 Dane
9.0 Norse

Distance to: Kevin_scaled
0.02570348 Swede:VK400
0.02602278 Dane: Denmark_FunenVK373
0.02772056 Saxon:I0777
0.02791175 Dane: Denmark_JutlandVK323
0.02929046 Celt:ISL_Viking_Age_Pre_ChristianSSG-A-4
0.02947827 Dane: Denmark_FunenVK370
0.02950685 Celt:England_IAI0156
0.02996588 Saxon:I0161
0.03018237 Celt:UK_Orkney_ScotlandVK202
0.03028748 Swede:VK34
0.03038180 Celt:ISL_Viking_Age_Pre_ChristianSTT-A-2
0.03152015 Swede:VK303
0.03153582 Dane: Denmark_FunenVK301
0.03322868 Dane: Denmark_JutlandVK300
0.03401761 Saxon:I0774
0.03478206 Dane: Denmark_FunenVK134
0.03481573 Dane: Denmark_JutlandVK325
0.03586989 Swede:VK33
0.03643067 Swede:VK402
0.03677194 Dane: Denmark_JutlandVK324
0.03691711 Swede:VK39
0.03746959 Norse:VK388
0.03792479 Dane: Denmark_LangelandVK275
0.03805439 Norse:VK414
0.03899464 Dane: Denmark_LangelandVK361

You left out your distances for your breakdown.

[gixajo]

It is just as bad as a underfitted model (with too high dist), the difference is that a overfitted model extracts noise from the data as if it represents real variation, while underfitted models fails to roughly corresponds to the data distribution. In a nMonte context the idea is to not use too many closely related populations, and in your model you are using 90 very closely related references, which makes it total crap in terms of statistical power.

She can correct that effect simply adding penalization.usually only with 0.25X results get better.

[gixajo]

You left out your distances for your breakdown.

I told him the same, but he added them quickly.

[Token]

I do understand that but being Irish these samples are realistic for my ancestry. A lot of the people that came were quite similar which explains why there hasn't been enormous changes since the Bronze Age but thanks for taking the time to explain this.

The percentages may be realistic by pure chance, it is possible that you are really 60% Celtic and etc. But the method that you used to reach these percentages is completely unreliable, so the final result is unreliable too.

[RyoHazuki]

Use my Iberian ancient sources, Galician reference you used is modern.

Target: Ryo_scaled
Distance: 1.6634% / 0.01663360
46.2 Celt
38.4 Swede
5.6 Iberia_Northeast_Empuries2
4.4 Iberia_Southeast_c.10-16CE
3.6 Iberia_Northeast_c.6-8CE_ES
1.8 Iberia_Southwest_EN

[Grace O'Malley]

The percentages may be realistic by pure chance, it is possible that you are really 60% Celtic and etc. But the method that you used to reach these percentages is completely unreliable, so the final result is unreliable too.

Someone else on Anthrogenica did the model. I mean I'm mostly Gael so that would be the majority of my ancestry.

[gixajo]

Someone else on Anthrogenica did the model. I mean I'm mostly Gael so that would be the majority of my ancestry.

Only for curiosity, could you add to those you posted, Iberian samples I posted?

I think that the this would add maybe some little % Iberian but will increase your Celt %.

Adding them, even my Andalusian father obtained near 20% Celt with aroun 1.0 distance.

[mitalit]

Target: Xabier_scaled
Distance: 2.2974% / 0.02297421
97.2 Gallo-Roman
2.8 Celt

Xabier_scaled
0.03619320 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR109
0.03876978 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR110
0.04158504 Gallo-Roman:FRA_IAJeb8
0.04237752 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR108
0.04257313 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR105
0.04480465 Gallo-Roman:FRA_IAERS88
0.04737335 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR33
0.04898363 Gallo-Roman:FRA_IANOR4
0.05052375 Gallo-Roman:FRA_IANOR3-15
0.05100550 Gallo-Roman:FRA_IANOR2B6
0.05106245 Gallo-Roman:FRA_IACOL153i
0.05459697 Gallo-Roman:FRA_IAERS1164
0.05628419 Gallo-Roman:FRA_IAATT26
0.05636661 Gallo-Roman:ITA_Rome_Late_AntiquityRMPR106
0.05936349 Gallo-Roman:FRA_IABFM265
0.06304029 Daneenmark_FunenVK301
0.06631633 Gallo-Roman:FRA_IAERS86
0.06689408 Saxon:I0161
0.06730064 Daneenmark_JutlandVK324
0.06772251 Swede:VK402
0.06821464 Gallo-Roman:FRA_IAPEY53
0.06835718 Celt:UK_Orkney_ScotlandVK202
0.06866284 Gallo-Roman:FRA_IACOL153A
0.06906821 Celt:England_IAI0156
0.07105677 Swede:VK39

Target: Xabier_scaled
Distance: 1.2709% / 0.01270877
40.6 Gallo-Roman
30.6 Iberia_Southeast_c.10-16CE
21.6 Celt
5.0 Iberia_Northeast_Empuries2
1.6 Iberia_Southeast_c.3-4CE
0.2 Iberia_North_BA
0.2 Iberia_Southeast_BA
0.2 Swede

[Token]

She can correct that effect simply adding penalization.usually only with 0.25X results get better.

She should use an entirely different software if she really wants to test very closely related populations, because this is not something that Global25 is made to do. Davidski himself says that, just like he repeatedly says that you shouldn't injudiciously add very closely related references to a model, but people apparently don't care about the basics and these crappy calculators keep proliferating like rats.