Comment by dbspin

That's an idiotic premise. There's no circumstance in which it wouldn't be better to load in the missing data computationally.