Cross validation for imputation

Question

[switch.reem]

Dear all

I am trying to perform cross validation on the data used for genome imputation. To be honest I have little knowledge on how to perform this and from the understanding of my research I will mask either the SNP or the genome. My question is this, I am using impute2 for imputation so do I do the data in the reference data or the gen file to be imputed? And the second question is how do I do the masking, do I remove the values or replace them with a zero? I would really appreciate if there is any paper or research that you can also point out that explains the process