Closed dllahr closed 5 years ago
This is a great addition! The tests failed because it looks like there's a stray pdb
in test_fast_corr.py
.
Thanks Lev! Should be fixed now. In other news, I used it to run a spearman calculation between ~17k vectors. When I did with a script to loop over them, it took ~10 days. I used nan_fast_spearman today and did it in less than an hour.
87% of the values have less than 10% discrepancy between the calculations.
(edited to use fractional difference instead of absolute difference)
Added methods to calculate covariance and correlations when there are nan values present - skipping the nans. Uses linear algebra and other numpy methods (instead of loops) to try to keep it as fast as possible.