Is there a way to analyze temporal (or spatial) point pattern using an inhomogeneous poisson process? I’m running into some issues because the poisson expect a y that is an integer, but I would like y to either be either a vector of time or a vector of location (x,y coordinates).
Can you say more about what you want to do, this isn’t making sense.
Use integer array instead of vector.
This StanCon 2017 notebook
http://mc-stan.org/events/stancon2017-notebooks/stancon2017-sanders-lei-mass-shootings.html
has Poisson model in time and declares integer observations as int z1[N1];
1 Like
I have series of event that occurs on a continuous time (point processes) that I would like to model using an inhomogeneous poisson process. So my response variables is a vector of time. I could use a vector of ones for the response vector (y) but then it seems that it would have hard time converging. I could also discretize the time, and look at how many events happened within 30minutes for example but then it seems that the way I discretize time will have a big effects on the outcome of the model.
A poisson process generates discrete events at points in time so the model
is either for event occurrence/count in which case you can use the poisson
pmf with discrete data or you may be measuring the inter-arrival times in
which case you would have continuous data but their distribution would most
definitely not be poisson so I think the ambiguity in your question remains
Thank you! the notebook is very useful. My issue is that my events are binary so my integer vector or array would just have a bunch of 1s. Theoretically, I could add a bunch of 0s for the rest of the time but I wouldn’t have any covariates associated with these times.