How to correctly scope a dataframe in R to use with a formula in a function

Question

After some digging (1, 2, 3), there appears to be a few posts about using formulas within functions causing scoping issues if I am understanding them correctly. Some suggests to use an environment, assign, or <<- to get around the issue, but I've been stumped how to use them (and confused why there's an issue in the first place).

Let's try this toy code:

library(survival)
library(survminer)
set.seed(1)

give_p_val <- function() {
  df <- data.frame('OS' = ovarian[, 'futime'], 'Survival_event' = ovarian[, 'fustat'])
  subgroup <- sample(nrow(df), nrow(df)/2)
  df$Class <- 'A'
  df$Class[subgroup] <- 'B'
  
  fit2 <- survfit(Surv(OS, Survival_event) ~ Class, data=df) 
  return(surv_pvalue(fit2))
}
give_p_val( )

It doesn't work, unless you run it directly, which hints at a scoping issue.

This code will work to return a fitted object:

survfit(Surv(futime, fustat) ~ rx, data=ovarian)

So why does the function break if we copy a dataframe within the scope?

testit<-function(){
  ovarian2 <- ovarian
  fit2 <- survfit(Surv(futime, fustat) ~ rx, data=ovarian2)
  return(surv_pvalue(fit2))
}
  
testit()

Ultimately, how do I generate a dataframe within a function to be handled correctly by the formula being used? Thanks!

Answer 1

This is a known issue in the survminer package and how surv_pvalue interacts with survfit objects. If you change survfit to the survminer package's version, surv_fit, your function will work just fine.

give_p_val <- function() {
  df <- data.frame('OS' = ovarian[, 'futime'], 'Survival_event' = ovarian[, 'fustat'])
  subgroup <- sample(nrow(df), nrow(df)/2)
  df$Class <- 'A'
  df$Class[subgroup] <- 'B'
  fit2 <- surv_fit(Surv(OS, Survival_event) ~ Class, data=df) 
  surv_pvalue(fit2, data = df)
}
give_p_val( )
#   variable      pval   method pval.txt
# 1    Class 0.2615363 Log-rank p = 0.26