Hi all,
I’m looking for advice (statistical and computational) to do something like the following simulation/inference for nonhomogeneous point processes (temporal, not considering space):
- Specify a “prior” for the intensity function as a spline or some other function for which it is easy to enforce constraints (e.g. cyclic over work weeks, monotonic increasing, etc)
- Simulate from the NHPP under the prior
- When data arrives, update posterior distribution, do more simulations until more data arrives.
I’m not particularly attached to any specific estimation method, in fact having something that could be run relatively fast (MAP, Laplace approximation, etc) would be more valuable than something that took a much longer time.
I hope the question isn’t too vague. Thanks for any papers, tutorials/examples, packages, etc that anyone can suggest!