CODEWITHSUNDEEP
pymc3
Andrew Nelson
Andrew Nelson

Reputation: 595

Uniform has no logp method

I must be missing something here. I can calculate the log probability from a Normal distribution, but not a Uniform distribution:

>>> import pymc as pm
>>> with pm.Model():
...     b = pm.Normal('b', 0, 1)
...
>>> b.logp({'b': 2)
array(-2.9189385332046727)

>>> with pm.Model():
...     a = pm.Uniform('a', 0, 1)
...
>>> a.logp({'a': 0.5})
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: 'TransformedRV' object has no attribute 'logp'

Upvotes: 1

Views: 706

Answers (1)

aloctavodia
aloctavodia

Reputation: 2070

PyMC3 transform bounded distributions, like the uniform, to unbounded ones. So you could ask PyMC3 to do not transform the variable:

with pm.Model() as model:
    a = pm.Uniform('a', 0, 1, transform=None)
a.logp({'a': 0.5})

or ask for the transformed variable like:

with pm.Model() as model:
    a = pm.Uniform('a', 0, 1)
model.logp({'a_interval__': 0.5})

You may want to check the unobserved_RVs and free_RVs attributes of model

Upvotes: 2

Related Questions