Fisher's exact test on rows in data frame - R

问题

I have a data frame of n rows that resembles below (with some extra columns containing additional information not listed):

           R1counti    R1counto     R2counti    R2counto    R1    R2
sample1    100         100          1000        1000        1     1
smaple2    50          100          50           50         0.5   1

For each row, I want to perform a fisher's exact test to determine if the R1 ratio is significantly different from the R2 ratios (and also in the end get an adjusted p-val)

Desired output (with the count columns still included in the out):

           R1    R2      pval
 sample1   1     1       1
 sample2   0.5   1       0.05

I thought I could do something like:

dataframe$p-val <- with(dataframe, p-val <- fisher.test(R1, R2, alternative= "two.sided"))

But I get an error:

FEXACT error 40.
Out of workspace.

Maybe I'm not handling this situation correctly. Any advice? Suggestions? Can I perform a fishers test like this?

回答1:

For each row of your dataset you need to build a matrix (i.e. a 2 x 2 contingency table) and pass this matrix to the fisher.test command.

df <- read.table(text="
          R1counti    R1counto     R2counti    R2counto    R1    R2
sample1    100         100          1000        1000        1     1
sample2    50          100          50           50         0.5   1
", header=T, row.names=1)


apply(df, 1, 
      function(x) {
          tbl <- matrix(as.numeric(x[1:4]), ncol=2, byrow=T)
          fisher.test(tbl, alternative="two.sided")$p.value
      })

#    sample1    sample2 
# 1.00000000 0.01209383

来源：https://stackoverflow.com/questions/52784396/fishers-exact-test-on-rows-in-data-frame-r