Kaspias
11-06-2018, 06:40 PM
At first thanks to randwulf and Davidski for all of their helps.
Modern
[1] "1. CLOSEST SINGLE ITEM DISTANCE%"
Italian_Abruzzo:ItalyAbruzzo17--> 2.064316
Bulgarian:Bulgaria1--> 2.101761
Turkish_Istanbul:Istanbul25095--> 2.104139
Macedonian:Macedonian2--> 2.121320
Serbian:Serbian_Serbia5--> 2.188790
Montenegrin:Montenegro1--> 2.189452
Montenegrin:Montenegro7--> 2.204881
Turkish_Balikesir:Balikesir16653--> 2.234278
So, it is a combination of Balkan, Turk, and Mediterranean types, which matches with my ancestry.
Run with all modern models;
1] "distance%=0.7262"
Kaspias
Ashkenazi_Jew,18.4
Montenegrin,15.4
Turkish_Balikesir,8.4
Turkish_Aydin,7.2
Italian_Abruzzo,5
Turkish_Istanbul,5
Macedonian,4.4
Serbian,4.2
Bulgarian,2.6
French,2.6
Assyrian,1.6
Turkish_Adana,1.6
Azeri_Iran,1.2
Bashkir,1.2
Bosnian,1.2
Cypriot,1.2
German,1.2
Hazara,1.2
Italian_Tuscan,1.2
Druze,0.8
Kazakh,0.8
Greek,0.6
Iranian_Persian,0.6
Turkmen,0.6
Uygur,0.6
Yukagir_Tundra,0.6
Austrian,0.4
Belgian,0.4
Iranian_Fars,0.4
Iranian_Zoroastrian,0.4
Itelmen,0.4
Mansi,0.4
Moldovan,0.4
Mordovian,0.4
Swedish,0.4
And this from Gradient Descent:
distance%: 0.35
Macedonian 17.23
Ashkenazi_Jew 13.99
German 12.44
Assyrian 12.12
Polish 11.9
Lebanese_Druze 9.11
Greek_Trabzon 5.02
Altaian 4.99
Eskimo_Chaplin 2.98
Selkup 2.54
Saudi 2.35
French 1.96
Cypriot 1.47
Bantu_Kenya 1.29
Itelmen 0.61
They both are "over fit", meaning that they are producing models that mathematically fit, but don't reveal the best model of the ancestry.
They do reveal some things, usually, though and give some starting point information. I see the Balkan, Mediterranean, and Turkic type of ancestry. There seems to be lots of East Asian/Siberian minor ancestry, which is typical of Slavic and Turkic ancestry. It is interesting that both like to use the Ashkenazi in the model at a high percentage. (I don't have known Jewish ancestor.)
So, I came up with this combination that is simpler, and seems accurate to me, i'm not sure though but this is best one that i can reach now. I am not sure if maybe the G25 spreadsheet references don't include enough of the right populations.
distance%: 0.88
Montenegrin 28.7 --> Balkan-Slavic
Cypriot 21.9 --> Eastern Mediterranean
Serbian 16.21 --> Balkan-Slavic
Turkish_Balikesir 13.79 --> Eastern Mediterranean - Turkic
Bashkir 10.35 --> Slavic-Turkic
Kazakh 9.05 --> Turkic
The Montenegrin and Serbian represent Balkan ancestry, the Turkish Balikesir, Bashkir, and Kazakh representing the more Turkic and Central Asian ancestry, and the Cypriot filling out the Eastern Mediterranean ancestry. However, there may be a better way to represent it. But this is the best one for now.
Also i want to show this one. Interesting, i'm not sure if does it mean anything though.
distance%: 3.81
Bulgarian 90
Hun-Sarmatian 10
Ancient
Well i tried to do something on ancient but i really need your advices. I stuck with ancient cultures.
Gradient Descent to analyze:
distance%: 1.1
Balkans_IA 21.38
Scythian_Moldova 21.09
Hasanlu_IA 20.98
Scythian_Ukraine 12.71
Hun-Sarmatian 8.0
England_IA 6.59
Saami_IA 3.23
Anatolia_IA 3.03
Scythian_Hungary 2.98
I tried to simplify this. This is the where i finally came, didn't work perfectly. But at least worked. (Btw i used 30% filter out deviating in this one, i got more accurate results with it.)
distance%: 2.48
Balkans_IA 42.5
Hasanlu_IA 22.5
Altai_IA 20
Baltic_IA 15
That's all for now.
Modern
[1] "1. CLOSEST SINGLE ITEM DISTANCE%"
Italian_Abruzzo:ItalyAbruzzo17--> 2.064316
Bulgarian:Bulgaria1--> 2.101761
Turkish_Istanbul:Istanbul25095--> 2.104139
Macedonian:Macedonian2--> 2.121320
Serbian:Serbian_Serbia5--> 2.188790
Montenegrin:Montenegro1--> 2.189452
Montenegrin:Montenegro7--> 2.204881
Turkish_Balikesir:Balikesir16653--> 2.234278
So, it is a combination of Balkan, Turk, and Mediterranean types, which matches with my ancestry.
Run with all modern models;
1] "distance%=0.7262"
Kaspias
Ashkenazi_Jew,18.4
Montenegrin,15.4
Turkish_Balikesir,8.4
Turkish_Aydin,7.2
Italian_Abruzzo,5
Turkish_Istanbul,5
Macedonian,4.4
Serbian,4.2
Bulgarian,2.6
French,2.6
Assyrian,1.6
Turkish_Adana,1.6
Azeri_Iran,1.2
Bashkir,1.2
Bosnian,1.2
Cypriot,1.2
German,1.2
Hazara,1.2
Italian_Tuscan,1.2
Druze,0.8
Kazakh,0.8
Greek,0.6
Iranian_Persian,0.6
Turkmen,0.6
Uygur,0.6
Yukagir_Tundra,0.6
Austrian,0.4
Belgian,0.4
Iranian_Fars,0.4
Iranian_Zoroastrian,0.4
Itelmen,0.4
Mansi,0.4
Moldovan,0.4
Mordovian,0.4
Swedish,0.4
And this from Gradient Descent:
distance%: 0.35
Macedonian 17.23
Ashkenazi_Jew 13.99
German 12.44
Assyrian 12.12
Polish 11.9
Lebanese_Druze 9.11
Greek_Trabzon 5.02
Altaian 4.99
Eskimo_Chaplin 2.98
Selkup 2.54
Saudi 2.35
French 1.96
Cypriot 1.47
Bantu_Kenya 1.29
Itelmen 0.61
They both are "over fit", meaning that they are producing models that mathematically fit, but don't reveal the best model of the ancestry.
They do reveal some things, usually, though and give some starting point information. I see the Balkan, Mediterranean, and Turkic type of ancestry. There seems to be lots of East Asian/Siberian minor ancestry, which is typical of Slavic and Turkic ancestry. It is interesting that both like to use the Ashkenazi in the model at a high percentage. (I don't have known Jewish ancestor.)
So, I came up with this combination that is simpler, and seems accurate to me, i'm not sure though but this is best one that i can reach now. I am not sure if maybe the G25 spreadsheet references don't include enough of the right populations.
distance%: 0.88
Montenegrin 28.7 --> Balkan-Slavic
Cypriot 21.9 --> Eastern Mediterranean
Serbian 16.21 --> Balkan-Slavic
Turkish_Balikesir 13.79 --> Eastern Mediterranean - Turkic
Bashkir 10.35 --> Slavic-Turkic
Kazakh 9.05 --> Turkic
The Montenegrin and Serbian represent Balkan ancestry, the Turkish Balikesir, Bashkir, and Kazakh representing the more Turkic and Central Asian ancestry, and the Cypriot filling out the Eastern Mediterranean ancestry. However, there may be a better way to represent it. But this is the best one for now.
Also i want to show this one. Interesting, i'm not sure if does it mean anything though.
distance%: 3.81
Bulgarian 90
Hun-Sarmatian 10
Ancient
Well i tried to do something on ancient but i really need your advices. I stuck with ancient cultures.
Gradient Descent to analyze:
distance%: 1.1
Balkans_IA 21.38
Scythian_Moldova 21.09
Hasanlu_IA 20.98
Scythian_Ukraine 12.71
Hun-Sarmatian 8.0
England_IA 6.59
Saami_IA 3.23
Anatolia_IA 3.03
Scythian_Hungary 2.98
I tried to simplify this. This is the where i finally came, didn't work perfectly. But at least worked. (Btw i used 30% filter out deviating in this one, i got more accurate results with it.)
distance%: 2.48
Balkans_IA 42.5
Hasanlu_IA 22.5
Altai_IA 20
Baltic_IA 15
That's all for now.