PDA

View Full Version : Modelling known ancestry with nMonte Runner (No OptimaFit)



BalkanKiwi
06-30-2020, 04:48 AM
GenoPlot is the first opportunity I've had to model +4 populations, which is somewhat exciting. After a bit of experimentation, its possible to over fit the models slightly by doing a number of things, such as replacing Croatian with Serbian, and Māori with Kosipe. I'm not overly concerned with Māori or Kosipe, because as mentioned in the OptimaFit thread, it seems most of my Polynesian is actually Papuan like. Here are 2 quick models I've come up with that show my known ancestry the best with a reasonable fit. The percentages for Croatian and Ashkenazi are somewhat off, but being so mixed I don't expect nMonte to work wonders in this regard. It does a pretty fine job of giving a ball park figure for each. Both of these are scaled with penalty removed, as this seems to produce more accurate results by not excluding the minor Oceanian (or inflating it). The second model is closer to the paper trail, as I can fit in Scottish and Welsh without losing Oceanian. Māori is removed when both are added. I believe this sample is not 100% Māori, so its possible its taken up by Scottish and/or Welsh (assuming they are mixed with Western European) when they are all in the same model. I'd need to play around with the sample to check this.

Model 1

https://i.imgur.com/3jLS6gW.jpg

Model 2

https://i.imgur.com/3x0CSZT.jpg