I did a log transformation on my observed variables before entering them into my model. Now I would like to take the exponential of them so that when I do things like have an az.plot_ppc plot in original variable rather than its log transformation. Any advice on how I can do that or if I even can?
You can use
InferenceData.map to apply the same transformation to both observed_data and posterior predictive (and or to prior predictive too) which are the quantities used in plot_ppc. This is the first example in arviz.InferenceData.map — ArviZ 0.14.0 documentation. Let me know if you have any issue and I’ll try to clarify both here and in the docs.
Thank you, this answers my question
I have found one more question. Do I have to make any changes if I take the exponent of my data? I seem to get a blank page once I map np.exp to it.
I think I have hit on the issue. My precited variables after the exponental transformation were larger in some cases than the observed data so I had to both resize the graphe axise as well as trim the observed data