Open GregSilverman opened 4 years ago
Hi, first of all, I have to thank you for doing this.
Glancing at the derivation you present for the expression used in f1_score_confidence_interval
everything looks copacetic. I won't have time to absolutely validate this by working it out by hand till after next week, but unless I missed something critical, I believe what you have is correct.
That being said, I think your output parameter names in your example run need to be switched, that is x_upper_bound
<-> x_lower_bound
(since your example upper bound is really the lower bound of the ci and vice versa)...
Thanks again!
Hi, I stumbled across this repo, since I needed a way to compute confidence intervals on F1-score for an NLP ensembling problem.
An initial glance at your derivation looks right. I'll let you know the outcome of my use of this.