I’m trying to fit an AR1 a time-series as following: with pm.Model() as ar1_mat1: theta = pm.Normal('theta',mu= 0, sigma = 3.0) tau = pm.HalfNormal('tau',sigma =1) likelihood = pm.AR1('likelihood', k=theta, tau_e = tau, observed = y_standard.values) trace = pm.sample(1000,…

If you want to generate posterior predictions “by hand”, then you don’t need sample_posterior_predictive(). You can just unpack parameter values from the posterior returned by pm.sample(). Then plug them back into your model as you describe. If you haven’t already, you should check out the AR1 no…

Pm.sample_posterior_predictive() from an AR1 process

Questions

alitirmizi January 24, 2022, 3:52pm 2

Also, is there a cleaner way to make future predictions using AR1/AR in pymc3?

Topic		Replies	Views
Silly question: 'zigzag' curves while making inference version agnostic arviz	3	520	April 8, 2020
Posterior predictive sampling with data variance Questions	10	2423	September 14, 2018
Trouble understanding sample_ppc Questions	4	651	June 7, 2018
Sample from prior? Questions	9	3525	July 20, 2018
Sample_posterior_predictive dependent variable vs deterministic dependent variable Questions	4	1129	February 11, 2020

Pm.sample_posterior_predictive() from an AR1 process

Related topics