ProbBins() and .get_probs()
This issue is to continue the discussion about creating an exported function GetProbs(). I've checked the two functions ProbBins() and .get_probs(), and they do use quantile() at some point but the following calculations are different.
This is what I understood:
-
ProbBins() computes probabilistic bins of a set of forecast years relative to the forecast climatology over a chosen period. It returns binary values indicating which category this forecast year belongs to. The output has member dimension because the function examines the category of each member in each forecast year.
-
.get_probs() creates probabilistic forecasts. It doesn't return member dimension since members are all together being calculated for the probability.
-
They both can specify the climatology period and can do cross-validation. .get_probs() includes weighting but ProbBins() doesn't. ProbBins() accepts the threshold to be either quantile (relative thresholds) or the absolute thresholds of the bins; .get_probs() accepts the relative thresholds (between 0 and 1) only.
I think we can create the big function GetProbs
because it produces the different result from ProbBins(), and the algorithm is different (at least seems to me; you know more than I do regarding this.) Please let me know your opinion, thanks!
Best,
An-Chi