For example,
ANALYZE <population> MODELS WHERE DEPENDENCE PROBABILITY OF X WITH Y > 0.5
ESTIMATE PROBABILITY OF x = 1 GIVEN (y = 2) WITHIN <population>
USING MODELS WHERE DEPENDENCE PROBABILITY OF x WITH y > 0.5
(From a note to myself. This is decidedly non-Bayesian, of course! It is also not even obviously a good idea even to show to non-Bayesians. Shooting this issue down is welcome but I figured it should be shot down explicitly rather than left to wither and die of forgotitude on its own.)