There is definatly a funnel there (between logQH_0 and mu_0), also some multi-modality in age_0.
I would first try more informative prior that cuts out the tail area, in GP you can almost never have too informative prior 
There is definatly a funnel there (between logQH_0 and mu_0), also some multi-modality in age_0.
I would first try more informative prior that cuts out the tail area, in GP you can almost never have too informative prior