Re-structure spatialstats.py

[rhugonnet]

The module spatialstats.py is starting to be a bit messy!

My thoughts right now:

• Move the spatial variogram scikit-gstat wrappers into a stats/spatial.py in GeoUtils,
• Move the N-D binning from scipy wrappers into a stats/binning.py in GeoUtils,
• Rename xdem/spatialstats into xdem/uncertainty.py? Move the fitting of a sum of variograms directly to scikit-gstat?

Any thoughts @adehecq @erikmannerfelt?

[rhugonnet]

From @erikmannerfelt's comment in #158:
Should also move _pandas_str_to_interval to misc.py

[adehecq]

Good point ! Yes, it can indeed be a bit difficult to find its way through all the functions... Some ideas/comments:

• I agree that the scikit-gstat wrappers could be moved to a separate (hidden?) submodule
• I'm also fine having a module named uncertainty.py containing the higher level functions highlighted in the examples. I don't know if I would move the binning into a separate module though. Wouldn't it be a very small submodule?
• after our discussions in PR Add BiasCorr classes and rename previous coreg.BiasCorr in coreg.VerticalShift #158, and my experience working with these tools yesterday: wouldn't it make sense to also have a binning and parametric option to calculate the heteroscedasticity? I was working on limited point observations, and in that case, the nearest neighbor extrapolation for values outside the domain didn't make so much sense, while a parametric fit would be super useful !
• I also noticed while going through the documentation that in the API, one cannot get access to the full docstring of functions in spatialstats.py (and other modules), only a summary. This forced me to check in the code directly, which is not so nice for beginners. Why is that?

[rhugonnet]

OK, perfect!
The binning stuff takes quite a bit of space, happy to add parametric fits (I also added a piecewise application of the original binning in spatialstats.py in #158).
Yes, the API of xDEM is not yet ready, we'll need to organize stuff a bit like in GeoUtils.