Hi everyone, I have a question that I can’t seem to easily answer.
Once I get the full posterior distribution of a parameter, let’s say \beta_1, during inference.
How do I get the density of the parameter being over 1, 2, 3 or some value ?
How do I easily tell and show that 20% of the prob. density is under 0.5 or over 1.3 for example?
Thanks
1 Like
Do you have the trace with the posterior samples? If so you can use numpy for simple stats
np.mean(trace['beta_1']>1.3)
4 Likes
As Nicholas said, it’s the beauty of Bayesian inference: you just have to count the scenarios of interest among all the possible scenarios in your model.
In addition to stats, you can use arviz.plot_posterior
's ref_val
argument: https://arviz-devs.github.io/arviz/generated/arviz.plot_posterior.html#arviz.plot_posterior
3 Likes
Exactly that, thank you!!
Good stuff, thank you guys for the help =)