avatar
Elliot Lipnowski @elliotlip.bsky.social

A tiny #mathsky note on conditional expectations. I'd previously seen two standard existence results of different strengths: a) For square-integrable RVs (L2) via a super easy Hilbert space argument. b) For integral RVs (L1, a bigger class) via Radon-Nikodym—harder! But (a) can actually prove (b)!

jul 22, 2025, 2:30 pm • 4 0

Replies

avatar
Elliot Lipnowski @elliotlip.bsky.social

It's nothing obscure—it's in Kallenberg. But I just saw for the first time today. - By orthogonal projection, an L2 RV x admits a conditional expectation T(x). - By Jensen or direct argument, T is (weakly) L1-contractive. - But L2 is dense in L1, so T extends to (bounded linear map on) L2.

jul 22, 2025, 2:30 pm • 2 0 • view