ahl27 / r_patches

Things I've done or am working on for base R

R Patches

Known bugs that I'm working on improving in R. Feel free to submit an issue on this repo and I'll add more improvements to my queue. This list is partly for my own reference so I don't forget what I was planning to work on.

Patches submitted

• #18480: Non-recursive reimplementation of dendrapply that also allows for different tree traversal methods.
• #18655: Improvements for *wilcox functions when populations are larger than 50, reduces memory scaling.

Known Issues

• *signrank functions hang for large n: Happens around n>=1000. Solution is likely to implement an exact=c(TRUE,FALSE) argument with a large population size approximation (e.g., Edgeworth).
• wilcox.test produces incorrect median estimates: $estimate is wildly incorrect in presence of zeros, ties, or degeneracy. This issue has been addressed and fixed in DescTools (link), but still needs to be fixed in base.
• wilcox.test inflates estimated p-values at tails: if either population is large (>=50), p-values will be inflated due to normal approximation. Edgeworth expansion should be implemented for improved accuracy.