Integration in joint likelihood

Modeling
1

I’m hoping to run MCMC using STAN (using the rstan package since I’m using R). My likelihood function involves an integral.

Is there a way I can program this self-defined likelihood in STAN? Thanks for your help.

2

Well, you would need some way of calculating the integral. It would depend on whether you could approximate it well enough so it depends on how nice the integral is.

1 Like
3

Thanks. Is there a way I can keep the integral intact and have the sampling done on it?

4

@randommm has a PR to (try to) handle something like this

However, you would currently have to #include your own C++ code to utilize it because it is not exposed to the Stan language yet. Even then, it would still depend on how amenable h() * g() is to the tanh-sinh transformation.