- Fixed a typo in Kendall's tau calculation for censored data that resulted in upper limit flags not being applied when evaluating whether Yi d< Yj (see Eq 27 of Isobe+ 1986, ApJ, 306, 490l). Identified by Bennett-Skinner.
pymckendall()is deprecated and will be removed after the 0.3.x series.- Added a warning when
Nboot=1. A user may provide this, intending to compute the correlation coefficient using manyNperturbrounds on the original dataset. However,Nboot=1does not use the input dataset as-is, but instead performs a single bootstrap with replacement. The warning clarifies thatNboot=Noneis the appropriate input when seeking to only use perturbation of points within the uncertainties.
- When computing Kendall's tau with censored data and bootstrapping, the x/y limit flags were not resampled. This is fixed and the limit flags are now properly resampled.
- Check x/y limit arrays to ensure they contain sensible values.
- Update assertions on array lengths to provide more helpful debugging information.
- fix description of
return_distbehavior in doctrsing forpymccorrelation - Amend changelog to fix release date for v0.2.2.
- Fix missing pass of x/y limits through
pymckendall()wrapper
- Pearson's r was incorrectly imported as spearman's rho. The import has been fixed.
- Add
setup.pyand package for pypi release.
- Rewrite random number generation to use numpy's newer Generator approach.
- Fix bug in bootstrapping that would cause final entry in list to be skipped.
Initial Release