Calculating Odds Ratio between multiple columns of a dataframe - r

Question

I have the following dataframe:
structure(list(id = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12,
13, 14, 15), var1 = c(1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1,
1, 1), var2 = c(0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1),
var3 = c(1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1), var4 = c(0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1), outcome = c(1,
1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1)), row.names = c(NA,
-15L), class = c("tbl_df", "tbl", "data.frame"))
I would like to arrange a script to calculate all possible odds ratio (using chi square), with 95% CI and p values, between all columns and the column outcome.
How can I do that?
I installed epitools but it seems that I need a 2x2 contingency table and I am not able to apply the function to columns of a dataframe

Answer 1

With mapply, you can use the fisher.test function, which doesn't fail when the odds ratio cannot be calculated.
mapply(fisher.test, x=data[, grep("var", names(data))], y=data[,"outcome"])
But the output is a 7x4 matrix which cannot be tidied into a nice format. However, we can use lapply to perform Fisher's test for each column and then tidy the results with the broom package.
library(broom)
cols <- df1[,grep("var", names(df1))]
res_list <- lapply(as.list(cols), function(x) fisher.test(x, y=df1$outcome))
do.call(rbind, lapply(res_list, broom::tidy))
# A tibble: 4 x 6
estimate p.value conf.low conf.high method alternative
<dbl> <dbl> <dbl> <dbl> <chr> <chr>
1 0 1 0 77.9 Fisher's Exact Test ~ two.sided
2 Inf 0.505 0.204 Inf Fisher's Exact Test ~ two.sided
3 2.13 0.608 0.160 37.2 Fisher's Exact Test ~ two.sided
4 Inf 0.505 0.204 Inf Fisher's Exact Test ~ two.sided
Or using dplyr with map, reshaping first and then splitting on the name.
library(dplyr)
df1 %>%
pivot_longer(cols=starts_with("var")) %>%
split(.$name) %>%
map(~fisher.test(x=.$value, y=.$outcome)) %>%
map(tidy) %>%
map_df(~as_tibble(.))
Data:
df1 <- structure(list(id = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12,
13, 14, 15), var1 = c(1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1,
1, 1), var2 = c(0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1),
var3 = c(1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1), var4 = c(0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1), outcome = c(1,
1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1)), row.names = c(NA,
-15L), class = c("tbl_df", "tbl", "data.frame"))

Answer 2

The following code performs the computations as described in the question but 3/4 give errors.
library(epitools)
cols <- grep("var", names(df1), value = TRUE)
res_list <- lapply(cols, function(v){
tbl <- table(df1[, c(v, "outcome")])
tryCatch(oddsratio(x = tbl), error = function(e) e)
})
ok <- !sapply(res_list, inherits, "error")
res_list[ok]
The errors are all this:
simpleError in uniroot(function(or) { 1 - midp(a1, a0, b1, b0, or)
- alpha/2}, interval = interval): f() values at end points not of opposite sign
which can be seen with
res_list[!ok]