Metric to show improvement of multivariate predictions

[hsun3163]

Need to find a way to quantify the improvement of power via heritability. The number of significant gene could works, but variance explained seems to be a better indicator.

Maybe utmost paper have talked about it, need to reread.

[gaow]

@hsun3163 you can definitely compute heritability by definition, then compare.

Another typical approach is to split data into train/test sets and compute the correlation coefficient between the estimated and true values in the test set, using estimates from the train set.