search

SolarArbiter / SolarArbiter.github.io

Contains the Solar Forecast Arbiter static website
https://solarforecastarbiter.org/
MIT License
9 stars 8 forks source link

fix probabilistic metrics indices #192

Closed wholmgren closed 3 years ago

wholmgren commented 3 years ago

Fixes inconsistencies regarding the index used for the time. Now time index is i throughout document. Unfortunately, Wilks 2019 uses the opposite convention in the Brier score section, which I think this was originally modeled on.

wholmgren commented 3 years ago

@dplarson can you review this? I think this cleans up the terminology and symbols in a way that is consistent with Wilks 2019 but with i and k reversed (pdf below for reference).

Chapter-9---Forecast-Verifica_2019_Statistical-Methods-in-the-Atmospheric-Sc.pdf

dplarson commented 3 years ago

@wholmgren I support these changes, though I did not generate the HTML to check the LaTeX rendering. The only thing I'm still a little iffy on is the use of o_i as the event outcome. Or to be clearer, I don't think there is a problem with using o_i, but rather I have wondered whether a more explicit indicator-like notation (e.g. 1_{x <= x_0}) would be better. However, the verbosity of such a notation may not actually provide a clearer understanding. Which is a long winded way to say "I'm good with the changes and feel free to ignore my other comments unless you strongly agree."

EDIT: Actually, please ignore my comments on alternatives to o_i given that Wilks 2019 uses o_i and following a good reference like Wilks 2019 trumps using an alternative that may or may not be clearer.

wholmgren commented 3 years ago

All good points. Thanks for the review @dplarson!