Returning p value and correlation coefficient from modified cor.test

Question

I am trying to run a modified version of the cor.test, using the following code which I found on another handy thread on stackoverflow:

cor_withN <- function(...) {
  res <- try(cor.test(...)$estimate, silent=TRUE)
  ifelse(class(res)=="try-error", NA, res)
}

I am using this with the running() command to perform a bunch of moving window correlations that contain some NAs. I would like to return at least the correlation coefficient and p-value (but if I can get the test statistic and df as well, that would be fantastic).

I have tried removing the $estimate part, in hopes that it would return everything I mentioned above, but I got a pretty ugly, incomplete output.

The code (without the running() command):

cor_withN2 <- function(...) { 
    res <- try(cor.test(...), silent=TRUE) 
    ifelse(class(res)=="try-error", NA, res) 
}
cor_withN2(x, y)
[[1]]
   t 
1.948752 

I would appreciate any suggestions on how this could be modified to return all of the stats that I'm looking for, or another approach to achieving this result.

Thank you!

Answer 1

Welcome to SO!

I think your problem is that ifelse() is designed for vectors of conditional expressions as well as of alternative values, while the latter are automatically and intransparently recycled/truncated to agree with the length of the condition (in your case, probably shortened to one, so that only the t-value is returned).

Consider rephrasing the last statement from your function along the lines of

if (class(res)=="try-error") NA else unclass(res)[c("estimate","p.value")]

...(or try to provide an executable example that comes with sample data).