Cross validation error Admixture

Question

[benhalimayossra]

Dear all,

I'm using the ADMIXTURE's cross-validation procedure I executed this instruction

# !/bin/sh
for K in 3 4 5 6 7 8
  do admixture --cv hapmap3.bed $K | tee log${K}.out
done

My pruned data set were generated with plink software using this parameter (--indep-pairwise 50 10 0.8) but when I checked with the values of CV error I found an unexpected increasing values.

Could you help me please

[Zev.Kronenberg]

Do you have a reason to expect that you have more than 3 populations?

[benhalimayossra]

Yes, my reason is to compare the North African against the European populations and Middle East so here, I expect that I have more than 3 ethnicities

[Zev.Kronenberg]

Try K=2,3,4. If three is the lowest than three is the best choice given the data and the model. Another option is filtering out rare and high frequency variants.