Hi ricardoV94, thanks for your time.
I am not exactly sure, the gaussian in the middle of the ppc (where the lot of the blue lines are) does not show everytime, (even when I run the same code multiple times) it sometimes shows, sometimes not. Maybe it is just some kind of a bug or whatever.
But nevertheless. The “posterior predictive mean gmm” as labeled on the bottom graph indicated with dashed line does follow the observation closely. The infered means of the two underlying gaussians, however, do not fit the data. (See figure 1, red dashed line - the gaussian mixture reconstructed from mcmc fit. It is not one gaussians, but two gaussians with very cloae means and almost 50:50 weights)
I tried to fit the gmm with sklearn and fit it almost perfectly (original mean: 100, fit by sklearn: 101 on average). I was able to reconstruct the gmm with sklearn more quickly and more precisely.
I am still wondering what am i doing wrong here, or whether im using mcmc methods for something that theyre nkt meant to be used for…
Thank you,
Andrej