Hi, I’m having a pesky problem with multi-indexing, which I never seem to be able to get right. I have a multi-level beta-binomial model that models demand of certain product styles (10) over 52 weeks. Each of the 10 styles has a different probability p for each of the 52 weeks, but I’d like to p…

Can confirm the error, but getting the shape to work is quite tricky. My suggestion is to flatten your predictors and observers into 10*52 vector - similar hierarchical model are much easier to handled and there are quite some example around.

Sample_ppc shape error with multi-index model

Questions

junpenglao August 10, 2018, 4:48pm 5

No problem! I filed a bug on Github: https://github.com/pymc-devs/pymc3/issues/3147

I like the 1d vector set up better, because it is also easier if you have missing data.

Topic		Replies	Views
Bayesian multilevel model indexing problems Questions	1	520	June 4, 2021
Shape error when sampling more than one value Questions	13	1907	December 20, 2018
Fitting multiple measurements, shape issues in sample_ppc Questions	4	1448	November 8, 2018
Index error in hierarchical model Questions	1	815	June 5, 2018
Form of index for multi-dimensional shape with observed data Questions	3	1260	May 6, 2020

Sample_ppc shape error with multi-index model

Related topics