https://github.com/pydata/pandas/issues/7517
As @mynegation notes, you can use Anaconda (or virtualenv).