Add Bayesian interface to Turing.jl

[RaphaelS1]

[sourish-cmi]

Do you want to add it to Turing? And then just call it in SurvivalAnalysis.jl?

Or do you want to develop it as part of SurvivalAnalysis.jl with Turing.jl?

[RaphaelS1]

Hey so that's an open question into what the integration effectively looks like. My current thinking is maybe a bridge package like SurvivalBayesian.jl which depends on both SurvivalAnalysis.jl and Turing.jl and basically is just glue code to build models

[sourish-cmi]

My only concern is that if we build too many packages, each package's maintenance could be a headache. So SurvivalBayesian.jl could be part of your SurvivalAnalysis.jl. Of course, it will depend on both SurvivalAnalysis.jl and Turing.jl. We can simply use include(SurvivalBayesian.jl) in the main module of SurvivalAnalysis.jl. Let me know your thought.

[RaphaelS1]

Agreed - I'll give some more background tomorrow and then summarise our chat here

[sourish-cmi]

Wonderful. Looking forward to our discussion tomorrow.

[sourish-cmi]

• Adding this discussion on discourse here.
• The paper on rstanrm package for Bayesian Survival Analysis in R is here.

[RaphaelS1]

Thanks also adding this discussion from discourse