Kevin Sheppard
Kevin Sheppard
Can only publish releases on pypi, not arbitrary patch levels. You can also install the nightly wheels. Search in issues for the location on Anaconda.org.
You can also download the zip from github, extract it, CD to where you extracted the code, and then enter `pip install .` This is the modern equivalent to calling...
You might need to use `pip install . --no-build-isolation` To use this you will need to have installed numpy, cython, scipy, pandas and setuptools-scm
You probably need `wheel` too.
Sounds like you need to clone the repo locally, then you can build from the clone. Seems to be less easy to build from a zip (or impossible now).
First, when you start `python` or `ipython` you _should not_ be in the git repo. If you are in the git repo, it will look in the `statsmodels` folder and...
Sounds solved. I'm going to close, but please reopen if the issue pops up again.
Not a bug, but a different implementation. SM uses i+1/n for i in range(n) and n = len(data).
See https://en.wikipedia.org/wiki/Q%E2%80%93Q_plot#Filliben's_estimate
We have stopped testing on x86 as it frequently causes edge failures that appear as if random.