[R] Fisher's Exact Test on each row of a data frame

row_fisher_df.R

#!/usr/bin/env Rscript

# fisher test on each row
row_fisher <- function(row, alt = 'two.sided', cnf = 0.95) {
  f <- fisher.test(matrix(row, nrow = 2), alternative = alt, conf.level = cnf)
  return(c(row,
           p_val = f$p.value,
           or = f$estimate[[1]],
           or_ll = f$conf.int[1],
           or_ul = f$conf.int[2]))
}

# generate sample data for test
test_df <- data.frame(matrix(sample.int(1000, size = 4000, replace = TRUE), ncol = 4))
colnames(test_df) <- c('a', 'b', 'c', 'd')

# run
p <- data.frame(t(apply(test_df, 1, row_fisher)))
print(p)

row_fisher_dt.R

#!/usr/bin/env Rscript
#
#               outcome
#              +       -
#          +-------+-------+
#        + |   a   |   b   | a + b
#  group   +-------+-------+
#        - |   c   |   d   | c + d
#          +-------+-------+
#            a + c   b + d
#

sapply(c('dplyr', 'snow'), function(p) require(p, character.only = TRUE))
cl <- makeCluster(parallel::detectCores(), type = 'SOCK')

# fisher test on each row
row_fisher <- function(row, alt = 'two.sided', cnf = 0.95) {
  f <- fisher.test(matrix(row, nrow = 2), alternative = alt, conf.level = cnf)
  return(c(row,
           p_val = f$p.value,
           or = f$estimate[[1]],
           or_ll = f$conf.int[1],
           or_ul = f$conf.int[2]))
}

# generate sample data for test
df_test <- tbl_df(matrix(sample.int(1000, size = 4000, replace = TRUE), ncol = 4)) %>%
       rename(a = V1, b = V2, c = V3, d = V4)

# run
p <- df_test %>%
       parApply(cl, ., 1, row_fisher) %>%
       t() %>%
       tbl_df()
print(p)

Hi Daichi,
Excited to read your R script for doing fisher test for dataframe. I am new to R, and wonder what is the difference between the two "row_fisher_dt.R" scripts you showed here. And another thing is what is the meaning of "sample.int(1000, size = 4000, replace = TRUE)", should I replace it with my table?

Thanks,
Xiangbin