How to deal with a variable size of the 'observed' variable

[junpenglao]

You might find this post helpful where I was trying to explain the mechanism of DensityDist.

Problem model definition PyMC3

Hi, I have a custom likelihood and I’m trying to determine the posterior probabilty for my parameters , given Uniform priors. I’ve written (trying to follow https://arxiv.org/pdf/1507.08050.pdf): with Model() as paper: #priors m = Uniform('m',0.,1.) q = Uniform('q',0.,5.) sigmay = Uniform('sigmay',0.,1.) # deterministic rule logEpeak_i=(q)+(m*logEiso) #likelihood definition def likelihoodMCMC(m,q,sigmay,logEiso,logEpeak_i,logErrEiso,logErrEpeak_i): return np.exp(0.5*n…