Function mixup between scipy.stats and wafo.stats

[morbult]

By calling a probability distribution function in scipy.stats, the corresponding function in wafo.stats is actually called.

I discovered this when using scipy.stats.exponweib.fit() (fitting a Weibull distribution to data) but the point is illustrated simpler using scipy.stats.norm.freeze()

import scipy.stats
scipy.stats.norm.freeze()
Out[2]: <scipy.stats._distn_infrastructure.rv_frozen at 0xb9a6b38>

import wafo
scipy.stats.norm.freeze()
Out[4]: <wafo.stats.estimation.rv_frozen at 0xecb72e8>

Tested with scipy 1.2.1 and 1.4.1

So far I have not found the results from wafo.stats and scipy.stats to differ much. But there is a different behaviour with warnings (at least for the function scipy.stats.exponweib.fit()) which can be confusing and annoying. It is also a potential cause of hard to find bugs, so I hope it can be fixed. (If the problem is even within wafo which I honestly don't know.)

[pbrod]

This was a feature of wafo version 0.11.
This is no longer an issue in the current version 1.0.0 available from https://github.com/wafo-project/pywafo.
The main benefit of using wafo.stats over scipy.stats is that it extends
scipy.stats with the possibility to use maximum spacing (MS)
estimator as an alternative to the maximum likelihood (ML) estimator and
the technique known as profile likelihood to construct better confidence intervals.
See https://en.wikipedia.org/wiki/Maximum_spacing_estimation and
https://journals.sagepub.com/doi/pdf/10.1177/1536867X0700700305

To use the extended functionality for parameter estimation you must call the dist.fit2 method
or alternatively use the FitDistribution class in wafo.stats.estimation.
To construct better confidence intervalls for the estimated distribution parameters, probability or quantile you must use the Profile-, ProfileProbability- or ProfileQuantile-class, respectively, in wafo.stats.estimation module.

[morbult]

Great! I installed from the current master and it's no longer an issue as you said. I'm happy to use scipy's fit function for now.

Out of curiosity, how was this accomplished in the previous version? I really don't understand how importing wafo can redirect my call scipy.stats to wafo.stats