@chartl perhaps my problem can be reformulated into something that is more appropriate. This is my first use of these techniques so I’m very green both using pymc3 and the underlying theories.
I’m trying to estimate the production rate of some manufacturing lines including information on the variability. Unfortunately I only know how many cartons made within an 8 hour block and sometimes multiple products can be made in those blocks.
Simply stated that’s:
hours = 8 = carton_sku1/carton_per_hour_sku1 + carton_sku2/carton_per_hour_sku2 …
carton_per_hour_skun being what I want and being normally distributed.
When I solve this, I take ‘hours’ being normally distributed with observed data of ‘8’ (whereas really it’s deterministic).
The end result is the correct mean for ‘carton_per_hour’ but all the variability is in the ‘hours’ variable.
Is there a way I can reframe this to get the desired result?