Jerry Johnson, pillar and chief guru of the Amateur Microwaves mailing list, posted a link to this paper, which is a method for calibrating noise figure without either a calibrated noise source or a calibrated attenuator. I'm not sure how well it does in the presence of various amounts of mismatch, but it's worth analyzing more carefully than I have time for at present.
So for your radio frequency delectation,
Cute trick. I'll try it out one of these times.
Cheers
Phil Hobbs