rajanil / fastStructure

A variational framework for inferring population structure from SNP genotype data.
MIT License
134 stars 50 forks source link

Ke and Koc inconsistency #25

Open lucius81 opened 9 years ago

lucius81 commented 9 years ago

Dear all, I am running fastSTRUCTURE on my data set, and all evidence (PCA, AMOVA) indicate that there is no genetic structure in my data. My concerning issue is that when i run fastSTRUCTURE with the simple prior, the model complexity that maximizes marginal likelihood (Ke) and the model components (Koc) used to explain structure in my data are both = 1, as expected. However, when i use the logistic prior, Koc is still 1, but the Ke value uprise to the highest number of K tested in a particular run. If I test from K1 to K5, then Ke=5, and so on.... This concerns me because as far as i understand, the paramater that tends to overestimate K when genetic structure is weak or difficult to determine is usually Koc. Therefore, i don't know how to justify my results. Any response will be really appreciated. Thank you vey much in advance. Luciano