Hi, I am trying to make a scenario where three uniformly distributed parameters need to be constrained. p1 = pm.Uniform(‘p1’,lower = 0, upper = 1) p2 = pm.Uniform(‘p2’,lower = 0, upper = 1) p3 = pm.Uniform(‘p3’,lower = 0, upper = 1) I want the constraint to be p1+p2+p3 = 1. How can I do it? p …

Adding constraint to the parameter

junpenglao June 22, 2018, 9:06pm 17

Absolutely. Dirichlet is much better tested with a stable internal transformation to make sure model is identifiable (ie no need to worry about specifying n-1 free variables)

Topic		Replies	Views
Sampling from simplex with inequality constraints Questions	7	2729	May 20, 2019
Which Distribution for Constrained Parameter vector Sum Questions	2	432	August 10, 2018
How does pymc3 implement the Dirichlet distribution? Questions	7	2665	June 5, 2019
How to define a custom likelihood on a bounded support (on a simplex) Questions	4	1602	August 29, 2019
Define constraint on the priors Questions	12	2147	April 19, 2018

Adding constraint to the parameter

Related topics