@junpenglao Thanks for your help! So when I call the pm.gp.MarginalSparse().marginallikelihod method it requires parameter ‘y’ to measure the fit of the GP. However, I am currently modelling fit with the linear submodel and therefore y in my parameterization is a pymc3.model.freeRV. Can one put random variables inside the marginallikelihod method as the parameter to model fit? Edit: I think I just found it… is_observed parameter sets y as an array of data or a RV.
1 Like