sgkit
sgkit copied to clipboard
VanRaden GRM
Fixes #874
I've tried to preemptively make this function suitable for additional estimators if/when required.
One point I'm unsure on is if we should have a default for ancestral_frequency
based on the sample population. It's common practice to use the sample population allele frequency as an estimate of the ancestral allele frequency although this is often inappropriate. So the question is whether it's best to force an explicit choice by the user, or default to the (potentially inappropriate) most common choice?
So the question is whether it's best to force an explicit choice by the user, or default to the (potentially inappropriate) most common choice?
I think it's best to force a choice in the first instance if there's any doubt. Otherwise you can get into nasty problems of backwards compabibility, if the default that you initially chose turns out to be not quite the thing you should have done.
Is this ready to be merged or are you planning further changes @timothymillar?
I think it's good to go, merging