2
$\begingroup$

New to Bayesian Modeling and the python library PYMC. Got some confusing result. How would an expert on Bayesian modeling interpret these graphs?

with BVAR_model:
    trace = pm.sample(chains=4, random_seed=rng)

az.plot_trace(trace)

enter image description here

After changing the number of lags, from 2 to 4, and applied differencing I got the following result. The right hand side looks better, but the left hand side confuses, strange distribution.

enter image description here

$\endgroup$

1 Answer 1

1
$\begingroup$

Those trace plots don't link right. They should look like “hairy, fat caterpillars” i.e. something like the below.

”Hairy, fat caterpillar” trace plot.

Moreover, they should heave the same means, while on your plots we can clearly see different chains getting stuck at different values. It might suggest that your model is misspecified and theatre are identifability issues, by the folk theorem of statistical computing.

$\endgroup$
1
  • $\begingroup$ identifiability could definitively be an issue. I extended the number of lags in this BVAR model and applied differencing. There should be a correlation between the two time series I'm modeling, but perhaps not a very strong one. $\endgroup$
    – Henri
    Oct 14, 2022 at 11:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.