odds ratio instead of p-value for categorical variables

[xinkai-zhou]

Dream package! Kudos to its author Kazuki.

Now, for summarizing binary variables stratified by another binary variable, is there a way to output odds ratios (see attached picture taken from Agresti's CDA book) instead of p-values?

I guess there's an easy tweak to achieve this. Can you point me to the right place to get started? Thank you.

[kaz-yos]

You can capture the output as a string matrix, and from there everything is under your control.

## Create a vector of OR as you like
orObject

## Capture output as a string matrix
tabAsStringMatrix <- print(TableOneObject, printToggle = FALSE, noSpaces = TRUE, test = FALSE)

## Add OR column
tab1 <- cbind(tabAsStringMatrix, orObject)

## print or write.csv()
tab1

[xinkai-zhou]

Thanks for your response, Kazuki.

I apologize for not making my points clear; here is what I meant to ask: I assume CreateTableOne() performs hypothesis testing in the background and extract p-value by fisher.test(data)$p.value; my question is whether it's possible to add the option of outputting fisher.test(data)$estimate, the conditional odds ratio, with some simple tweak to the CreateTableOne() function.

The solution you suggested is indeed elegant and doesn't require much work. But with CreateTableOne(), I can probably do some customization and streamline the process.

Thank you.

[kaz-yos]

Testing occurs with this helper function, which runs a given test function and extracts the $p.value part of the results. https://github.com/kaz-yos/tableone/blob/42382213fd27416c6bfe7a534d30fca25fd83cd2/R/modules-constructors.R#L239-L263

It's very hacky, but you can create a custom fisher.test function that returns the OR as the "p-value", and use this function instead of fisher.test when using CreateTableOne. I haven't tested it, so let me know how it goes.

my_or_fisher <- function(...) {
  out <- fisher.test(...)
  out$p.value <- out$estimate
  out
}