100%!

My additional recommendation for studies like this would be to also do it with cluster robust errors and worry a lot if they're giving you a much more conservative result than HLM. Different versions of this problem can crop up quite easily.
#statsmeme

Comments