Observation(s) the model PREDICTS, recomputed for every proposed theta. Defaults to source, which is right whenever the data enters from outside — a runtime-bound array or an empirical measure — because then the same observable serves both sides. Name it separately when the recording is itself declared in the forward pipeline: measurement noise applied to both sides is the same deterministic draw twice, so it cancels in the residual and the inference silently fits noiseless data.
name: predicteddescription:'Observation(s) the model PREDICTS, recomputed for every proposed theta. Defaults to `source`, which is right whenever the data enters from outside — a runtime-bound array or an empirical measure — because then the same observable serves both sides. Name it separately when the recording is itself declared in the forward pipeline: measurement noise applied to both sides is the same deterministic draw twice, so it cancels in the residual and the inference silently fits noiseless data.'from_schema: https://w3id.org/tvborank:1000owner: Likelihooddomain_of:- Likelihoodrange: stringmultivalued:true