I am doing genomic predictions of breeding values using a large dataset. I have around 10000 cows genotyped for HD (500K) and phenotyped for a range of traits. I am evaluating the accuracy of prediction based on the Pearson correlation between adjusted phenotypes and genomic breeding values (GEBVs) divided by the heritability for the trait. I also want to present the regression coefficients of adjusted phenotypes on GEBVs. However, for traits with an accuracy of 0.53, I am finding an regression coefficient of 1.21 and for a trait with accuracy 0.3, I found a regression coefficient of 0.37, which is way too far from 1, indicating that my predictions are too biased. Should I divide the regression coefficient by the heritability as well? I have checked everything and I cant find any error. Any help/comments, please?
Thank you very much.