Summary to pandas dataframe

numpyro
1

Dear all,
Thanks for the nice package.
Is there anyway to convert print_summary() to pandas data frame.
I can make an arviz object but for large problems it takes actually longer than my sampling :slight_smile:.
Thanks,
Mahdi

2

@mahdik You can use summary utility for samples from mcmc.get_samples(group_by_chain=True) to get a dict of stats. Then you can use some pandas functions to convert that dict to a data frame I guess.

3

Actually, numpyro.diagnostics.summary(mcmc.get_samples(group_by_chain=True)) took as long as using arviz.from_numpyro(mcmc). While mcmc.print_summary() is printing almost instantly.
Thanks,
MK

4

That’s strange. print_summary called summary under the hood.