significance tests

Question

Opened this issue 6 years ago · 3 comments

For our statistics functions, like weighted_cor it would be nice to enable computing significance metrics (p-value...).

@bradyrx, perhaps you can help with this?

Answer 1 · 2019-03-09T23:11:50.000Z

@matt-long, yes I'd be happy to open a PR on this. Are you trying to keep esmlab slim on dependencies?

xskillscore has a pearson_r_p_value function that is based in numpy and written as an xarray ufunc. Although this only allows a two-tailed test.

I can also port over some functions from climpred (https://github.com/bradyrx/climpred/blob/master/climpred/stats.py) which accounts for autocorrelation in the time series by computing the effective sample size (Bretherton et al. 1999).

Answer 2 · 2019-03-11T20:06:15.000Z

I am fine with dependencies as long as they are good ones.

Answer 3 · 2019-03-28T18:50:55.000Z

Implemented in #90, although should stay open until effective sample size is incorporated (will get on this when time allows).