In nimble we allow determ nodes to be in givenNodes in getConditionallyIndepSets.
When I originally translated getConditionallyIndepSets for nimbleModel, I left out processing related to determ nodes, thinking that we could just pass through those nodes. I think this works in most cases, but if givenNodes includes any deterministic nodes, then it breaks.
I could add the additional processing related to determ nodes that is in nimble, but that will involve some additional work and I think it would mean getConditionallyIndepSets would be slower as we would be working through the graph using immediate=TRUE.
It's not clear to me if we have a use case for determ nodes as givenNodes and from a probabilistic standpoint, it's not clear what it means to condition on deterministic nodes. So, @perrydv , let's discuss before I do further work on this.
In nimble we allow determ nodes to be in
givenNodesingetConditionallyIndepSets.When I originally translated
getConditionallyIndepSetsfor nimbleModel, I left out processing related to determ nodes, thinking that we could just pass through those nodes. I think this works in most cases, but ifgivenNodesincludes any deterministic nodes, then it breaks.I could add the additional processing related to determ nodes that is in
nimble, but that will involve some additional work and I think it would meangetConditionallyIndepSetswould be slower as we would be working through the graph usingimmediate=TRUE.It's not clear to me if we have a use case for determ nodes as
givenNodesand from a probabilistic standpoint, it's not clear what it means to condition on deterministic nodes. So, @perrydv , let's discuss before I do further work on this.