I am a researcher at Oxford / Imperial, who is working to developer a suite of tools for inference on (hard) ODE and PDE models (see PINTS). Example models used by our collaborators include ODEs and PDEs used to model electrical conductance in heart cells or batteries.
We want to include general state-space modelling for the error component of the models, that is,
$y_t = f(\theta\; t) + \epsilon_t$,
where $f(\theta\; t)$ is a fixed deterministic mean and $\epsilon_t$ is the error.
Rather than reinvent the wheel, it'd be great to use another package, like yours potentially. Can your framework handle models like this? If so, given a value of $\theta$ (and hence all $f(\theta\; t)$ are known) and parameter values for whatever state-space error process, can the framework be used to calculate a likelihood?
Hi,
I am a researcher at Oxford / Imperial, who is working to developer a suite of tools for inference on (hard) ODE and PDE models (see PINTS). Example models used by our collaborators include ODEs and PDEs used to model electrical conductance in heart cells or batteries.
We want to include general state-space modelling for the error component of the models, that is,
$y_t = f(\theta\; t) + \epsilon_t$,
where $f(\theta\; t)$ is a fixed deterministic mean and $\epsilon_t$ is the error.
Rather than reinvent the wheel, it'd be great to use another package, like yours potentially. Can your framework handle models like this? If so, given a value of $\theta$ (and hence all $f(\theta\; t)$ are known) and parameter values for whatever state-space error process, can the framework be used to calculate a likelihood?