1
![Not allowed!](images/buttons/up_dis.png)
Thumbs Up |
Received: 2,275 Given: 2,559 |
That's super cool. They all look like much better distances then probably most of the people here would be getting I don't know it's frustrating because there needs to be more central European in this calculator and Germany is lacking as well as the Czech Republic and Austria and Hungary Swiss peoples and then I'm sure there's other groups I can't even think of but anyway it's the calculator sort of a screw up it needs more samples in my opinion to be correct at least throughout Central Europe probably large parts of France as well anyway live and learn
Thumbs Up |
Received: 11,047 Given: 26,801 |
use experimental mode for better distances:
https://vahaduo.github.io/chebyshev/
Thumbs Up |
Received: 2,275 Given: 2,559 |
I'll check it out thanks for pointing that out and do you know how accurate the custom PCA is on the official site everyone uses because I found out some interesting distances that shock me on that site you should try plotting some of your results into that site together with your own and see what you plot out with it's pretty interesting I think especially when you aggregate groups like Russia Smolensk is closer to me than Ireland in some ways although they both are almost identically on top of me with a couple samples but I don't know it seems strange to me
Thumbs Up |
Received: 2,275 Given: 2,559 |
Target: Brad_AncestryDNA_Scaled_Official
Chebyshev distance: 0.6512% / 0.00651206
49.4 Piedmont
18.0 Spain_Medieval
13.8 Estonia_Medieval.SG
8.6 Oland
3.8 Italy_Medieval_EarlyModern.SG
1.8 Turkey_Medieval
1.6 Pakistan_RajaGira_Medieval
1.4 Pakistan_Singoor_Medieval
1.2 Pakistan_Barikot_Medieval_2
0.4 Funen
Chebyshev dist: Brad_AncestryDNA_Scaled_Official
0.01239192 Piedmont:Italy_North_EarlyMedieval_Langobards_1:CL 146
0.01443272 Oland:Sweden_Viking.SG:VK337_noUDG.SG
0.01739108 Funenenmark_Viking.SG:VK373_noUDG.SG
0.02015474 Langelandenmark_Viking.SG:VK275_noUDG.SG
0.02130208 Nordic:Faroes_EarlyModern.SG:VK241_noUDG.SG
0.02769108 Oxford:England_Viking.SG:VK177_noUDG.SG
0.03029908 Funenenmark_Viking.SG:VK278_noUDG.SG
0.03042908 Villa_Magna:Italy_Medieval_EarlyModern_oCentralEur opean.SG:R62.SG
0.03399316 Russia_Sunghir_Medieval.SG:Sunghir6_noUDG.SG
0.03523188 Ukraine_Medieval.SG:VK541_noUDG.SG
0.04225016 Estonia_Medieval.SG:s19_IVLS09KT_1_noUDG.SG
0.04388222 Spain_Medieval:I10897
0.04411486 Estonia_Medieval.SG:s19_IIa_1_noUDG.SG
0.04527816 Estonia_Medieval.SG:s19_IIf_1_noUDG.SG
0.04955246 Spain_Medieval:I10853
0.05245946 Spain_Medieval:I10895
0.05568946 Spain_Medieval:I10892
0.05730446 Spain_Medieval:I10852
0.07305044 Russia_Caucasus_Medieval:Anapa-9.SG
0.07316914 Russia_Caucasus_Medieval:Anapa-10.SG
0.07926846 Italy_Medieval_EarlyModern.SG:R60.SG
0.07958014 Russia_Caucasus_Medieval:Anapa-11.SG
0.07991446 Italy_Medieval.SG:VK536_noUDG.SG
0.11352114 Syria_TellMasaikh_Medieval.SG:R3359.SG
0.11503014 Turkey_PostMedieval:I4539
Thumbs Up |
Received: 2,275 Given: 2,559 |
Target: PARENT.1.IMPUTED.DNAGEN.RSIDMERGED.OFFICIAL.G25
Chebyshev distance: 0.7008% / 0.00700809 | ADC: 0.5x RC
45.0 Oland
23.8 Piedmont
16.8 Russia_Sunghir_Medieval.SG
7.4 Syria_TellMasaikh_Medieval.SG
5.2 Nordic
1.8 Spain_Medieval
Target: PARENT.2.IMPUTED.DNAGEN.RSIDMERGED.OFFICIAL.G25
Chebyshev distance: 0.8637% / 0.00863733 | ADC: 0.5x RC
73.8 Piedmont
16.2 Oland
5.6 Spain_Medieval
3.0 Italy_Medieval_EarlyModern.SG
1.4 Syria_TellMasaikh_Medieval.SG
Brad.DNAGen.MrgImp
Chebyshev distance: 0.7150% / 0.00715049 | ADC: 0.5x RC
61.8 Piedmont
12.4 Oland
7.4 Russia_Sunghir_Medieval.SG
6.4 Funen
4.6 Spain_Medieval
3.8 Syria_TellMasaikh_Medieval.SG
2.6 Estonia_Medieval.SG
0.6 Nordic
0.4 Italy_Medieval_EarlyModern.SG
Last edited by ScandinavianCelt; 04-30-2024 at 12:09 AM.
Thumbs Up |
Received: 11,047 Given: 26,801 |
I guess PCAs work differently, each of the 25 components (coordinates) have different/unequal weights, if i understand it correctly, whereas in the normal vahaduo algorithms (both classic and experimental), the 25 components/coordinates have equal weights in the calculation.
Also important:
https://bioturing.medium.com/princip...y-894e8f6f4bfbKeep in mind that PCs are ranked by how much they describe the data. PC1 reveals the most variation, while PC2 reveals the second most variation. Therefore, differences among clusters along PC1 axis are actually larger than the similar-looking distances along PC2 axis.
Thumbs Up |
Received: 2,275 Given: 2,559 |
This is regular vahaduo R3P Modern Scaled G25 with my official Ancestry coords vs Chebyshev:
Target: Brad_AncestryDNA_Scaled_Official
Distance: 0.4716% / 0.00471591 | R3P
64.4 Dutch
29.4 Orcadian
6.2 Jat_Uttar_Pradesh
This is with chebyshev experimental distance:
Target: Brad_AncestryDNA_Scaled_Official
Chebyshev distance: 0.1797% / 0.00179733 | R3P
47.2 Orcadian
45.0 Swedish
7.8 Pamiri_Ishkashim
Is the chebyshev better?
Thumbs Up |
Received: 7,660 Given: 4,850 |
Target: Gallop_scaled
Distance: 2.1020% / 0.02101973
91.8 Spain_Medieval
4.2 Oxford
4.0 Piedmont
Target: Father_scaled
Distance: 2.7049% / 0.02704943
95.6 Spain_Medieval
3.6 Estonia_Medieval.SG
0.8 Oxford
Target: brother_scaled
Distance: 2.5669% / 0.02566878
100.0 Spain_Medieval
https://www.yfull.com/tree/E-BY7449/
E-V22 - E-BY7449 - E-BY7566 - E-FT155550
According to oral family tradition E-FT155550 comes from a deserter of Napoleon's troops (1808-1813) who stayed in Spain and changed his surname.
Thumbs Up |
Received: 11,047 Given: 26,801 |
Intuitively I would say the algorithm seems to work better on chebyshev mode, than it does with normal. It detects for example your swedish ancestry, which normal mode doesn't. Same happens with me, I always get more relevant groups (given my ancestry) on experimental.
The way I see it, experimental:
1-gives lower distances
2-yields better results (more realistic ancestry sources in relation to your real ancestry).
Thumbs Up |
Received: 14,142 Given: 11,908 |
R3P standard mode:
Target: Dušan_scaled
Distance: 2.4736% / 0.02473616 | R3P
60.6 Russia_Sunghir_Medieval.SG
30.4 Italy_Medieval_EarlyModern.SG
9.0 Spain_Medieval
R3P Chebyshev mode:
Target: Dušan_scaled
Chebyshev distance: 1.2632% / 0.01263202 | R3P
44.6 Russia_Sunghir_Medieval.SG
43.0 Spain_Medieval
12.4 Ukraine_Medieval.SG
🔴
🔵
⚪
There are currently 1 users browsing this thread. (0 members and 1 guests)
Bookmarks