I wanted to know if it is possible to compute the inverse CDF of a PYMC distribution.
This can be done with scipy.stats distributions simply by calling the ppf function, for example, with a scipy.stats.skewnormal we can have:
I am wondering if a similar implementation is available for PYMC distribution. For example for a
pm.SkewNormal with arbitrary parameters (mean, scale, and skewness); how can we compute the inverse CDF?
We haven’t implemented icdf for most distributions, but we have the infrastructure inplace now (it was added for sampling from truncated distributions): pymc/truncated.py at 7503730dd20d4e7318b31a9834951aae647929d7 · pymc-devs/pymc · GitHub
We should open an issue in our repository to start adding some of the most useful ones. I think we only have for geometric and normal so far.
Hmm, ok, do you reckon stats.skewnorm is the same distribution as pymc.SkewNormal ? I need to comput the ppf and I’m wondering if I could use the function that’s already implemented in scipy.
Do you need the function inside a PyMC model? In that case you cannot use it directly. Otherwise, yes I think it’s the same looking at the source code.