Open andLaing opened 6 years ago
Sounds like a good idea to me
And to me! I will take a look next week (about to travel now), but I would say it follows THE GREAT RULE, as we were told by THE GREAT MASTER, before he ascended to the Polish Heavens: eg, thous shall not reinvent the wheel!
I currently have a bandwidth of about 3 bits per hour, so I can't really look at it in any depth, but I think that this is a great idea in principle. It does mention other packages that 'might suit your needs better' so we'd better have a look at those too.
Hullo again, your friendly neighbourhood pedant here...
I've been having a look at the uncertainties package: https://pythonhosted.org/uncertainties/ which despite the info on that page is available from conda: https://anaconda.org/conda-forge/uncertainties
It seems a well written piece of code which would give us access to a standard way to calculate errors in all analyses, avoiding the dangers of everyone defining their own propagation formulae (always dangerous) and depending on their level of pedantry omitting terms or not....
It provides functions to define via covariance matrices and wrappers to let user defined/third party functions work with the value+error variables it provides and it plays well with numpy.
Thoughts?