-
Notifications
You must be signed in to change notification settings - Fork 0
Closed
Description
Some components of a causal problem require effect handlers to be applied to the Graph.model, whilst others do not. Common examples are the do operation, or a conditional expectation.
At present, _CPComponents have a way to track which effect handlers should be applied, but these are not respected in the current CausalProblem.lagrangian function.
What needs to happen is that lagrangian needs to determine how many different models it needs to create (as per the number of different combinations of effect handlers), sample from all of these, then pass the correct samples to each constraint and the causal estimand.
Don't forget that a component that has no effect handlers applied beforehand needs to use the Graph.model as it is.
Metadata
Metadata
Assignees
Labels
No labels