Which is the best dataset for an anova; raw or blup values?

Question

I have a set of 150 rice varieties, which have different environmental classifications. I have measured a trait, y, on four repeats of each variety. I want to test the effec