Partsworth estimation causes "bad energy" warning on some observations

Questions
8

Hi Dan, happy to hear you managed to make it work! Hope the predictions will make sense :wink:
Regarding your questions:

  1. Not sure I undestood what you mean by “slicing”, but you can give ArviZ coords argument to plot_trace to select only some parameters in the trace – note that this takes xarrays in the background, which can take a little time to get used to. What I do in this case is az.from_pymc3(my_trace) to understand the structure of the different chains and then use coords.
    Easier and useful could also be to just pass compact=True to az.plot_trace – this will plot your multidimensional parameters (e.g a 6-dimensional beta) into the same plot, allowing matplotlib to diplay it. You can also choose only the parameters you’re interested in with the vars argument.
    Just FYI, ArviZ developers are working on integrating Bokeh as backend for plotting, in addition to matplotlib… Stay tuned :wink:

  2. Exactly. More precisely, you have the pm.Data class now, which is neat and easy to use – see this example NB.

Hope this helps, and good luck with your model :vulcan_salute: