Thank you very much for the insightful reply! The external input of the system is a time series which does not have any analytical solution. Indeed, it corresponds to user input of a device. I want to do system identification where I fit model parameters given the system response to user input, the system being a set of differential equations.
The use case here is pharmacology where I want to model individual patient response by measuring amount of drug given and effect. I do have population parameter distribution which serve as priors and I want to use patient specific response to obtain more accurate individual parameter estimations.
I saw that scipy’s odeint is completely numerical so I didn’t anticipate that pymc needs analytical gradients. Do you think there is a way I could get this working with arbitrary input?
I couldn’t find any example online showing how to use the ‘t’ parameter in this context.
Best regards,
Jona