Gedmatch Oracle Poll

[Graham]

What in your opinion, is the best Oracle to use?

Oracle
Oracle-4
Oracle X

Will create a poll. You are welcome to compare.

[Damião de Góis]

First option by some distance

[Graham]

First option by some distance

It's the least shaky and risky. I like Oracle-4. I think that's only because my results don't run all over the place like some. Oracle X is the worst.

[Tropico]

Oracle X for me personally was more accurate than 23andme, McDonald and DNA tribes.

[Tropico]

Example :
Eurogenes EUtest V2 K15 Oracle-x Population Fitting
0 Unable to determine 0.01%
1 Spanish_Galicia 43.35%
2 North_Amerindian 14.83%
3 Spanish_Extremadura 8.70%
4 Bantu_S.E. 8.62%
5 Algerian 5.98%
6 Mayan 5.29%
7 Portuguese 4.68%
8 French 4.38%
9 Orcadian 4.15%

Eurogenes EUtest V2 K15 4-Ancestors Oracle
1 Mozabite_Berber + North_Amerindian + Southwest_English + Spanish_Galicia @ 9.283

Eurogenes EUtest V2 K15 Oracle results:
1 80.5% Spanish_Galicia + 19.5% Mayan @ 10.96

Oracle X was spot on. The other two were "ehh".

[Damião de Góis]

It's the least shaky and risky. I like Oracle-4. I think that's only because my results don't run all over the place like some. Oracle X is the worst.

I haven't explored Oracle-4 too much since i don't care much about the N population aproximation, like:


Using 4 populations approximation:
1 North_Italian + Spanish_Castilla_Y_Leon + Spanish_Castilla_Y_Leon + Spanish_Cataluna @ 2.293

But for single population it seems to be the same as Oracle.

[Graham]

I guess Oracle on it's own, one population wouldn't work for people like yourself Tropico. Mixed of different origins. But for someone like Alex DeLarge who's family, the majority have probably lived in Portugal or Iberia for thousands of years. One option would be suitable.

[Graham]

1 North_Italian + Spanish_Castilla_Y_Leon + Spanish_Castilla_Y_Leon + Spanish_Cataluna @ 2.293
The way I do that, would be like this. It wouldn't match you still.

http://imageshack.us/a/img28/9362/7sx8.gif

[Damião de Góis]

The way I do that, would be like this. It wouldn't match you still.

http://imageshack.us/a/img28/9362/7sx8.gif

Good thinking, but shouldn't X be at the center? :p
Anyway, that was K13 Eurogenes.

[Graham]

Good thinking, but shouldn't X be at the center? :p
Anyway, that was K13 Eurogenes.
Nah because you got Leon twice.

[Graham]

It's hard to know when you get results like..

Eurogenes K13
Oracle-4
1 West_Scottish + West_Scottish + West_Scottish + West_Scottish @ 3.999
Oracle
1 West_Scottish 3.3

Eurogenes K15
Oracle-4
1 Orcadian + Orcadian + Orcadian + Orcadian @ 4.637
Oracle
1 Orcadian 4.08

Dodecad V3
Oracle-4
1 Irish + Irish + Irish + Irish @ 6.207
Oracle
1 Irish 5.53

[Damião de Góis]

Nah because you got Leon twice.

The spanish regions that were introduced add to the confusion in my opinion, because they are all very castillianized, and are all very similar to each other.
So in that sense, EU Test should be good because it doesn't have them. Oracle 4 results seem balanced in my opinion for 1 population:


Using 1 population approximation:
1 PT @ 4.134
2 ES @ 5.224
3 North_Italian @ 8.362
4 FR @ 8.806
5 Tuscan @ 14.625
6 AT @ 17.347
7 Cornish @ 18.816
8 West_&_Central_German @ 19.078
9 NL @ 19.577
10 RO @ 19.854

But for 4 populations i would be all over the place using your method:


Using 4 populations approximation:
1 ES + FR + North_Italian + PT @ 2.828
2 FR + North_Italian + PT + PT @ 2.914
3 ES + ES + FR + North_Italian @ 3.054
4 ES + North_Italian + PT + PT @ 3.060
5 ES + ES + North_Italian + PT @ 3.141
6 North_Italian + PT + PT + PT @ 3.269
7 ES + ES + ES + North_Italian @ 3.491
8 ES + ES + ES + RO @ 3.499
9 ES + FR + North_Italian + North_Italian @ 3.501
10 ES + ES + FR + Tuscan @ 3.565

[Graham]

Hmm perhaps Polako should use contrasting Iberian regions for his Oracle. Whatever that is. Iberians would know the most polar-opposite regions from the lot.

[Damião de Góis]

Hmm perhaps Polako should use contrasting Iberian regions for his Oracle. Whatever that is. Iberians would know the most polar-opposite regions from the lot.

I think that would only be possible in the 15th century or so. Would be interesting actually.

[Graham]

I think that would only be possible in the 15th century or so. Would be interesting actually.

I ran K15 in a PCA with the Iberian regions plus neighbours if you are interested. Murcia pull away. I take it they have some outside influence or none?

http://imageshack.us/a/img854/4777/lb3b.png

[Argang]

In the 49-cluster run Polako did some Murcians formed a cluster with Canary Islanders and Cubans. Other Iberians formed a single cluster, which most likely means that the various regions of Spain and Portuguese are very similar. Tuscans and North Italians, for example formed two clusters .

[Damião de Góis]

I ran K15 in a PCA with the Iberian regions plus neighbours if you are interested. Murcia pull away. I take it they have some outside influence or none?


It's strange, they score the same as everyone else from what i've seen.

[Graham]

In the 49-cluster run Polako did some Murcians formed a cluster with Canary Islanders and Cubans. Other Iberians formed a single cluster, which most likely means that the various regions of Spain and Portuguese are very similar. Tuscans and North Italians, for example formed two clusters .
ok I looked it up from Forumbiodiversity. Because that is Mr Aussie Polakos home. Don't know what that's based off but it looks interesting.

Polako:
According to Day Tripper, who can't log in because his IP seems to be banned, the West Eurasian MDS16 is the best out of all the West Eurasian and European runs so far.
DayTripper:
And here's a cluster summary:

V1 = Cuban/Spanish Murcia/Canary Islands
V2 = North African Jews
V3 = Georgian/Abkhasian
V4 = North Ossetian/Balkar/Adyei
V5 = Armenian/Turkish
V6 = Ashkenazi Jews
V7 = Iraqi/Iranian/Assyrian
V8 = Lebanese/Cyprian/Syrian
V9 = French
V10 = Hungarian/German/Austrian
V11 = Iranian/Kurdish
V12 = South Italian/Greek
V13 = Lezgin/Kumyk
V14 = French Basque
V15 = Bedouin
V16 = Saudi
V17 = Saudi/Bedouin/Syrian/Yemeni
V18 = Egyptian/Bedouin/Yemeni
V19 = Jordanian/Palestinian/Bedouin
V20 = German
V21 = Polish/West Ukrainian
V22 = Russian/Belorussian/Northeast Ukrainian
V23 = Lithuanian
V24 = Romanian/Bulgarian/Balkans
V25 = Irish/Scottish/West English/Orcadian
V26 = Kind of a NW European/American cluster
V27 = SE English/British/American
V28 = Swedish/Norwegian
V29 = Chechen
V30 = Chuvash
V31 = Druze
V32 = A few more Druze
V33 = East Finnish/Finnish
V34 = Southwest Finnish/Finnish
V35 = Kargopol Russian/Erzya/Moksha
V36 = Estonian
V37 = Southwest Finnish/Finland Swedish
V38 = Spanish/Portuguese
V39 = Spanish Basque (Pais Vasco)
V40 = Tuscan/North Italian
V41 = Turkish
V42 = Tuscan
V43 = Mari
V44 = North African/Moroccan
V45 = Mozabite Berber
V46 = Nogay
V47 = Palestinian
V48 = Sardinian
V49 = Yemenite Jewish

[Caismeachd]

I'm never please with my results on any of them really.

Oracle
1 98.8% Dutch + 1.2% Mayan @ 1.51

Oracle 4
1 50% German +50% Orcadian @ 1.724
1 50% German +25% Norwegian +25% Southeast_English @ 1.394
German + German + Norwegian + Southeast_English @ 1.394

Oracle X

1 Dutch 64.00%
2 Norwegian 19.27%
3 Danish 4.08%
4 Ashkenazi 4.07%
5 Southeast_English 3.43%
6 Spanish_Valencia 2.75%
7 Karitiana 1.10%
8 Swedish 1.10%
9 Spanish_Cataluna 0.21%
10 Lebanese_Druze 0.00%

[Graham]

btw I done that PCA again but set everything that was below one to zero. Took at as noise. It's amazing how much pull some noise can make, if you don't anything proper to use.

I'm wandering how effect noise has on the Oracle.

http://imageshack.us/a/img838/2146/0th7.png

[Damião de Góis]

^ that actually makes a lot of sense geographically and even historically: Galicia, Leon, Portugal and Extremadura in a cluster, with Leon pulling more towards other spanish regions.

[ZephyrousMandaru]

ok I looked it up from Forumbiodiversity. Because that is Mr Aussie Polakos home. Don't know what that's based off but it looks interesting.

Polako:
According to Day Tripper, who can't log in because his IP seems to be banned, the West Eurasian MDS16 is the best out of all the West Eurasian and European runs so far.
DayTripper:
And here's a cluster summary:

V1 = Cuban/Spanish Murcia/Canary Islands
V2 = North African Jews
V3 = Georgian/Abkhasian
V4 = North Ossetian/Balkar/Adyei
V5 = Armenian/Turkish
V6 = Ashkenazi Jews
V7 = Iraqi/Iranian/Assyrian
V8 = Lebanese/Cyprian/Syrian
V9 = French
V10 = Hungarian/German/Austrian
V11 = Iranian/Kurdish
V12 = South Italian/Greek
V13 = Lezgin/Kumyk
V14 = French Basque
V15 = Bedouin
V16 = Saudi
V17 = Saudi/Bedouin/Syrian/Yemeni
V18 = Egyptian/Bedouin/Yemeni
V19 = Jordanian/Palestinian/Bedouin
V20 = German
V21 = Polish/West Ukrainian
V22 = Russian/Belorussian/Northeast Ukrainian
V23 = Lithuanian
V24 = Romanian/Bulgarian/Balkans
V25 = Irish/Scottish/West English/Orcadian
V26 = Kind of a NW European/American cluster
V27 = SE English/British/American
V28 = Swedish/Norwegian
V29 = Chechen
V30 = Chuvash
V31 = Druze
V32 = A few more Druze
V33 = East Finnish/Finnish
V34 = Southwest Finnish/Finnish
V35 = Kargopol Russian/Erzya/Moksha
V36 = Estonian
V37 = Southwest Finnish/Finland Swedish
V38 = Spanish/Portuguese
V39 = Spanish Basque (Pais Vasco)
V40 = Tuscan/North Italian
V41 = Turkish
V42 = Tuscan
V43 = Mari
V44 = North African/Moroccan
V45 = Mozabite Berber
V46 = Nogay
V47 = Palestinian
V48 = Sardinian
V49 = Yemenite Jewish

Hey, I know this is probably too much to ask. But do you think you can run an MDS using all the populations from that analysis?

[Graham]

Hey, I know this is probably too much to ask. But do you think you can run an MDS using all the populations from that analysis?

I'll give it a shot. I rekon it will look messy though.

[Graham]

Am trying, it's taking a long time. If it crashes. Will take out people and try again tomorrow.

[Atlantic Islander]

Oracle-4 is the funniest.

[paksaltopam]

Definitely NOT Oracle-X. It has me as Bulgarian or Romanian every time. Sometimes Ashkenazi too.

[Atlantic Islander]

Oracle-4 is the funniest.

Although I have to say that Eurogenes Oracle-4 is fine (for the most part), I think it's MDLP and Dodecad that tend to be weird.

[Atlantic Islander]

Yes :nod:
EUtest 4-Ancestors Oracle
1 ES + PT + PT + PT @ 3.767
Dodecad World9 4-Ancestors Oracle
1 Galicia + Galicia + Galicia + Murcia @ 0.811
Dodecad K12b 4-Ancestors Oracle
1 Castilla_Y_Leon + Extremadura + Galicia + Galicia @ 1.619

Eh :shrug:
EUtest V2 K15 4-Ancestors Oracle
1 Spanish_Galicia + Spanish_Murcia + Spanish_Murcia + Spanish_Murcia @ 4.431
Eurogenes K13 4-Ancestors Oracle
1 Spanish_Castilla_Y_Leon + Spanish_Murcia + Spanish_Murcia + Spanish_Murcia @ 1.784
Dodecad K7b 4-Ancestors Oracle
1 Andalucia + Canarias + Spaniards + Valencia @ 0.354

No :no000000:
Dodecad V3 4-Ancestors Oracle
1 Morocco_Jews + Portuguese + Spanish + Swedish @ 1.452
HarappaWorld 4-Ancestors Oracle
1 basque + morocco-jew + spaniard + ukranian @ 2.786
MDLP K=12 4-Ancestors Oracle
1 CITAL + GER + IBR + PRT @ 2.759
MDLP World 4-Ancestors Oracle
1 Ire8 + Serbian + Sephardim + Sardinian @ 0.802
MDLP World-22 4-Ancestors Oracle
1 Portuguese + Russian_North + Sardinian + Sicilian @ 1.749

[Mortimer]

Oracle 4

[Graham]

Hey, I know this is probably too much to ask. But do you think you can run an MDS using all the populations from that analysis?
Couldn't do it too hard.. Tried a run map for you with population matches of Assyrians. And cut out most overfits.

http://imageshack.us/a/img15/2172/tv2p.png

[Stormer99]

Although I have to say that Eurogenes Oracle-4 is fine (for the most part), I think it's MDLP and Dodecad that tend to be weird.

My results seem to be similar in some aspects from calculator to calculator. I get Samaritan on every calculator and types of Italian on the calculators. Oracle 4 works better for people who aren't just one ethnicity.

[Atlantic Islander]

Oracle 4 works better for people who aren't just one ethnicity.

Yeah, seems so.

[Stanley]

My results seem to be similar in some aspects from calculator to calculator. I get Samaritan on every calculator and types of Italian on the calculators. Oracle 4 works better for people who aren't just one ethnicity.

Yeah, but even so, Oracle-4 can be a little screwy.

For instance, Eurogenes can't seem to decide whether I'm Germanic or not:

K15 closest match: Dutch + German + Norwegian + Southwest_English @ 4.021
K13 closest match: French_Basque + Irish + Irish + Ukrainian_Northeast @ 4.223

:lol00001:

[Stormer99]

Yeah, but even so, Oracle-4 can be a little screwy.

For instance, Eurogenes can't seem to decide whether I'm Germanic or not:

K15 closest match: Dutch + German + Norwegian + Southwest_English @ 4.021
K13 closest match: French_Basque + Irish + Irish + Ukrainian_Northeast @ 4.223

:lol00001:

But what if there are consistencies you notice in all the oracle 4 results? My oracle 4 results tend to be pretty consistent from test to test, especially the Italian and Samaritan. If you get something on one test, you are unlikely to have this sort of ancestry. If you get it on all tests, you either have it or have ancestry from a similar source population.
I see you getting Irish, British, and a type of Scandinavian for instance so it's not too inaccurate.

[Stanley]

But what if there are consistencies you notice in all the oracle 4 results? My oracle 4 results tend to be pretty consistent from test to test, especially the Italian and Samaritan. If you get something on one test, you are unlikely to have this sort of ancestry. If you get it on all tests, you either have it or have ancestry from a similar source population.
I see you getting Irish, British, and a type of Scandinavian for instance so it's not too inaccurate.

I'm sure for some people it's rather accurate. You might be one of those people. It's just that in my case it's too inconsistent and doesn't resemble my ancestry all that much.

My other Eurogenes Oracle-4 top results:
French_Basque + NO + NO + South_&_Central_Swedish @ 3.739 (EUtest)
FR + IE + IE + NO @ 3.107 (Jtest) -> this one probably makes the most sense

Then there's the absurdity from Dodecad and MDLP, which I won't even bother posting here.

Then again, I can't really say the other oracle options are any better; my specific mix of ancestries might just be bad for this sort of thing.

[Atlantic Islander]

Then there's the absurdity from Dodecad and MDLP, which I won't even bother posting here.


Oh man, way to make a person curious lol.