Daniel Turek

Sounds reasonable to me. This just means we could not use getParam() for distribution parameters of dimension > 2, right? But custom distribution functions could still be defined with parameters...

Sounds reasonable to me.

Briefly - this is not a bug. NIMBLE's handling of NA's in multivariate nodes is only written to sample the entirely multivariate node (consisting of entirely NA's). There is no...

Useful, for certain, but my concern is that some of the conditional distributions (up to a constant of proportionality) may be difficult to derive analytical expressions for, in the non-normal...

TJ, my apologies for the delay in responding. I took a look over your code, and it's a nice approach to sampling the missing dimensions of multivariate nodes. Thanks for...

I think this is worthy of group discussion at some point. I agree, it needs attention.

@perrydv Can we merge this PR (1) into devel, and (2) then into `ADoak` ? That would be my preference, as soon as we're ready, so I can move forward...

@perrydv Also, what would you think about changing the name of `abstract` (in the new `methodControl` list) to something more intuitive? Something that conveys whether a default (do-nothing) implementation will...

@perrydv Along the same lines, it also feels like a "double-negative" having to provide `abstract = FALSE`, in order to specify that I **do** want a default to be provided.

@perrydv Also (sorry for all the comments), why don't we just make **all** methods non-abstract, for **all** `nimbleFunctionVirtual` methods? (and skip the distinction between two different types of abstract /...