A mistake in the initial design of the Stan-aware prediction stack resulted in two separate branches of Experiment classes---one Stan-aware (ILkhdExperiment and descendants) and one not (IChiSqExperiment and descendants). This wreaks havoc on MultiExperiment, which needs fairly complex wrapping (see https://github.com/chenel/lblpwgtools/pull/1) to support it.
It's possible instead to consolidate them back into the same class and simplify the inheritance tree. This PR proposes how to do that.
A mistake in the initial design of the Stan-aware prediction stack resulted in two separate branches of Experiment classes---one Stan-aware (ILkhdExperiment and descendants) and one not (IChiSqExperiment and descendants). This wreaks havoc on MultiExperiment, which needs fairly complex wrapping (see https://github.com/chenel/lblpwgtools/pull/1) to support it.
It's possible instead to consolidate them back into the same class and simplify the inheritance tree. This PR proposes how to do that.