Wrong formula for Jaccard index?

[luciacsep]

🐛 Bug

When using the repOverlap function using the Jaccard index, I think it calculates wrong numbers. When looking at the source code in overlap.R, I saw that the Jaccard index is calculated as follows:

jaccard_index.default <- function(.x, .y) {
  .x = collect(.x, n = Inf)
  .y = collect(.y, n = Inf)
  intersection = nrow(dplyr::intersect(.x, .y))
  intersection/(nrow(.x) + nrow(.y) + intersection)
}

However, doesn't the intersection need to be subtracted:
intersection/(nrow(.x) + nrow(.y) - intersection)
?

I'm using Immunarch v. 0.5.5.

Thank you.

[vadimnazarov]

Hi @luciacsep ,

thank you a lot for noticing this critical issue! It will be fixed in the next release.

[vadimnazarov]

Hi @luciacsep !

Thank you again for noticing the bug, we published immunarch on CRAN and moved it to GitHub, feel free to install the latest versions and check if everything is OK for you. I'm closing this issue for now, please feel free to reopen it in case of more related questions.