Different SD between pm.summary() and trace.posterior.std()

[Dirk_Nachbar1]

I have a normally distributed parameter and pm.summary() reports a smaller sd then when I use trace.posterior[‘’].std(axis=(0, 1)) (2 chains).

Does pm.summary() not pool across chains when calculating SD?

[daniel-saunders-phil]

Hello, thanks for you question. Could you post a code snippet that reproduces your problem? For example, I tried:

with pm.Model():
    pm.Normal('m')
    trace = pm.sample()
pm.summary(trace)

and

trace.posterior['m'].std(axis=(0, 1))

But I found identical results.

[Dirk_Nachbar1]

my fault, I must have run the colab cells at different times, issue can be closed

[daniel-saunders-phil]

ah no worries - glad you sorted it out.