How to use geom_signif from ggpubr with a chi-square test-CodePudding

I am trying to use geom_signif(), from the package ggpubr. I am attempting to use it to do a Chisq test on my data (here is a sample) :

df_test= data.frame(cat = c("A","D","R"), count1 = c(10,12,23),count2 = c(9,3,4))

I am trying to see if there are some categories that are statistically enriched in certain categories in count2, which is a subset of count1. I've not seen an example online of geom_signif being used for another test than the wilcoxon, so I don't know the correct syntax to tell it what my variables are. here is the one I attempted :

melt_count = data.table::melt(df_test, id.vars = "cat")
ggplot(melt_count, aes( x =cat,fill=variable,  y=value)) geom_bar(stat = "identity", position = "dodge") 
  geom_signif(comparisons = list(c("A","D"),c("count1","count2")), test = "chisq.test")

Warning : Computation failed in stat_signif(): missing value where TRUE / FALSE is required

I know this mostly doable by doing the test outside of ggpubr but I specifically want to use it, or another ggplot-affiliated package able to do statistical test by itself.

Thanks in advance,

CodePudding user response：

The problem is that the test argument requires a function that takes two vectors to compare, whereas chisq.test takes a single matrix. We could therefore create a little wrapper function:

chi.test <- function(a, b) {
  return(chisq.test(cbind(a, b)))
}

ggplot(melt_count, aes( x =cat,y=value)) 
  geom_bar(aes(fill=variable), stat = "identity", position = "dodge") 
  geom_signif(comparisons = list(c("A","D")), 
              test = "chi.test")

^{Created on 2022-06-14 by the reprex package (v2.0.1)}