Hey, Alex,
please excuse my disappearance.
I changed the complete cleaning process. Now it seems to work 
Next month I get new data, which I use to test how well my attempt with Pymc3 works
However, I still have two questions, which were also not quite clear to me yet.
Is it possible to slice the trace plots somehow? I usually have a lot of partworths/beta weights and matplotlib doesn’t want to display them all at the same time…
Is it possible to output the traceplots with another visualization framework (e.g. plotly)?
To make a prediction with my model I just have to define it as theano.shared, correct?
For example, would my choices have to be created as a shared variable?
best regards
Dan