Hi, im building a model of sleep stage transitions. I’ve successfully implemented a Markov chain model where the next sleep stage is dependent on the previous. If the current_stage is either wake=0 or sleep=1 then the next stage can be categorically sampled from a transition matrix. For example: …

Just for anyone who comes across this in the future, here is the solution that worked: with pm.Model() as model: BoundedLower = pm.Bound(pm.Normal, lower=0, upper=1) alpha = BoundedLower('alpha', mu=priors['alpha'], sd=1e5, shape=(nst…

I believe i’ve fixed my own issues. I used a softmax to force the contraint that rows of p_trans sum to one. I also realized i need better priors on alpha, so i don’t get wild aphas in the first place. I’ll post the complete solution here when I fix the kinks.

Awesome, i’ll cheeck it out :smile:

Hazard Rates for Markov Chain Model

Questions

junpenglao May 25, 2019, 6:02am 3

You might find this discussion here helpful when it comes to puting softmax in your model: Multinomial hierarchical regression with multiple observations per group (“Bad energy issue”)

Topic		Replies	Views
Custom Categorical Distribution - ensure bounded candidates v5 modeling	13	496	March 31, 2024
Adding Conditional Transition Probabilities to a Hidden Markov Model version agnostic	5	207	November 18, 2024
Hierarchical HMM with dependent transition matrices	7	159	October 14, 2025
Posterior to Prior in Categorical distribution (or encapsulating multiple data sources to categorical analysis) Questions modeling	4	1841	February 28, 2022
Extreme running time for my graduation project v5 modeling	4	135	June 27, 2025

Hazard Rates for Markov Chain Model

Related topics