barakbri/hhg

Dev version of the HHG CRAN package

HHG

This is the GitHub page for the HHG package for non parametric statistical tests of independence and equality of distributions (K-Sample tests).

Installation

For latest CRAN (stable build), run:

install.packages("HHG")

For lastest development version run:

install.packages("devtools")
devtools::install_github('barakbri/HHG', build_vignettes = TRUE)

hhg's Issues

Pairwise test for a group of distributions

Hi there,

Imagine you have N different distributions and want to do pairwise HHG test of dependence on them. If the N is not too big, you can do N(N-1)/2 pairwise tests—but it is be challenging if there are many distributions to be compared. I needed to do this for N=50 (making 1225 pairs; DOI:10.31234/osf.io/EAGZD) and had to put it on a cluster.

I haven't looked under the hood of the hhg.test() function yet. I can imagine one can at least avoid partitioning over and over if the internal objects were passed along the results. Then, another function could take over and do the pairwise comparisons.

Is that—or something similar—reasonable/feasible? How hard is implementing it?

Thanks in advance!