Function for Impulse Response Function

Question

I created following example code to draw an "impulse response function" in R with help of the [vars][1] package.

library(vars)
data(Canada)
Canada <- data.frame(Canada)

irfplot = function(x, y) {
  VAR <- VAR(cbind(x,y), p = 2, type = "trend")
  irf_o <-irf(VAR, impulse = colnames(VAR$y)[1], response = colnames(VAR$y)[2], boot = TRUE, cumulative = FALSE, n.ahead = 20, ci = 0.90)
  plot(irf_o)
}

irfplot(Canada["rw"],Canada["U"])

That should work so far. However, when trying to make the script more flexible by writing the function as

irfplot = function(x, y, lags, deter) {
      VAR <- VAR(cbind(x,y), p = lags, type = deter)
...

irfplot(Canada["rw"],Canada["U"], 2, "trend")

it returns:

Error in VAR(y = ysampled, p = lags, type = "trend") : 
  object 'lags' not found 

Question: How can the issue be resolved? I have some other functions that transfer values via objects, but for any reason, it does not work.

Thank you.

Answer 1

Issue

The issue is the boot = TRUE argument to the irf() function. First note that, the following works just fine

irfplot <- function(x, y, lags, deter) {

  var_o   <- VAR(cbind(x, y), p = lags, type = deter)
  irf_o <- irf(var_o, 
               impulse = colnames(var_o$y)[1], 
               response = colnames(var_o$y)[2], boot = FALSE)
  plot(irf_o)
}
irfplot(Canada["rw"], Canada["U"], 3, "trend")

Changing boot to TRUE causes the error. What happens is that lags and deter apparaently do not get passed on correctly to the the function that performs bootstrapping. Although, I dont think its technically a bug, it would certainly be helpful if the author of the package changed it.

Solution

Whenever you want to pass on some function arguments in your top-level function to some lower level function its less error prone (as you see in your example) and generally recommended to use the ... argument.

irfplot <- function(x, y, ...) {

  var_o   <- VAR(cbind(x, y), ...)
  irf_o <- irf(var_o, 
               impulse = colnames(var_o$y)[1], 
               response = colnames(var_o$y)[2], boot = TRUE)
  plot(irf_o)
}
irfplot(Canada["rw"], Canada["U"], 3, "trend")