Hi Pymc3 Community, I have a data generating process where I am trying to model contaminates. My observed data is generally sampled from a single Normal distribution but occasionally is sampled from a contaminate process (which is modeled with a number of Uniform Distributions). I’d like to infer …

Definitely option 2, but with a Beta prior for theta. You can use something like a \text{Beta}(1, 10) to reflect the strong sparseness.

Best way to model a observed variable that is a mixture of two distributions

bdyetton November 19, 2018, 7:58pm 3

Awesome stuff! Thanks.

Topic		Replies	Views
The results of a switch are observed directly (beginner) Questions	1	924	November 11, 2018
Implementation Questions: Thompson Sampling and Switching Models Development	3	2275	December 30, 2017
Mixture of continuous and discrete logp Questions	6	1607	December 15, 2018
Condition Categorical Variable on Bernoulli Parents Questions	16	2340	May 14, 2019
Switching between class distributions Questions	2	787	January 12, 2020