In general I found it is much easier to start with a very simple (or even simplistic) model. There is a great lure to write down a complicated model that takes care of all the different effects you can think of, but at first I never manage to come up with something that is actually reasonable (I make that mistake a lot 
). There just are so many different small things that you never now why the model doesn’t work. Usually the sampler just doesn’t like it (and often for good reasons, the model might just be terribly miss-specified). If you build up the model peace by peace, then at least you know which new part introduced the problems. That also makes it easier to get help. As it is, it is just a lot of work just to understand what your model is doing in the fist place.
1 Like