I have a question about binary time series modeling. I have longitudinal data (binary credit risk) for a number of individuals. I have tried different GLM models but in vain. I would like to try modeling by using Gaussian Process. Unfortunately the example in stan guide here contains an example where time series is not considered explicitly.
I stumbled upon the sentence:
The Gaussian processes chapter presents Gaussian processes, which may also be used for time-series (and spatial) data.
What is implied by this sentence?
A simple approach would be adding time as covariate. I wonder if there are approaches to deal with time explicitly? A complicated factor is that there is periodicity (incentive is applied in a given quarter, it is applied every second quarter for all individuals) so I thought about one GP when incentive is not applied (baseline GP) and two GPs when incentive is applied (baseline+incentive GP).
Thanks for any advice