I am studying a python library that uses Bayesian inference to identify the coefficient of a linear regression.
I have two questions, one very broad and one more vertical on MCMC and numpyro.
- What is the sampling trace of a Bayesian Model? From my understanding It seems to be the mean of the distribution once I reached the convergence, but I am not sure about that.
In the library specifically here the trace is defined as
self.trace = mcmc.get_samples()
And this for me is slightly confusing from my first interpretation of "Trace", which makes me think I am wrong.
This because the mcmc.get_samples() from Numpyro Documentation here
it seems a way to draw samples (I am not understanding how many, but this is another question off-topic)
If my question format doesn't respect Stack Exchange best practices, please just let me know.
I also read this very complete answer, but I am still struggling understanding the example with Pebbles and Trace
ps I am sorry but I don't have enough reputation to add as new tags numpyro and lightweightmmm
