Well, after messing with it for several days I've decided that even if I can prove my contention that the optimal stratification has the variance inversely proportional to the number of blocks, it's still pretty bogus because the variance changes with each query.
However, in the process of working through all that, I did formalize my approach to where it's very easy to justify stratifying it that way even without the proof.
I may still have something on the optimal number of strata, and that would actually be the more useful result.
No comments:
Post a Comment