Writing your own functions

Before you start: single-value decisions.
Setup: No data files are needed. Run setup and code blocks from the top of this page.
Your check: Define a function’s inputs, default arguments and returned value. Practise in the unit summary.

You have already used many functions. Now it is time to write your own. A function can accept one or more arguments, listed in parentheses. Its body, enclosed in curly braces, contains the operations to perform. For example, the mean of a numeric vector without missing values is its sum divided by its length. We can write sum(x) / length(x) as a function:

function(x){sum(x)/length(x)}

To access this function later, you can also save it as an object in your environment:

my.mean <- function(x){sum(x)/length(x)}

For an initial test, create a vector X (uppercase), call my.mean(), and compare the result with mean():

X <- c(50:100)

my.mean(X)
mean(X)

Add a default argument

A default supplies a value when the caller omits an argument. Start with a function that adds two numbers:

my_function <- function(x, y = 10) {
  result <- x + y
  return(result)
}
my_function(5)
my_function(5, 20)

[1] 15
[1] 25

x receives 5. In the first call, y uses its default of 10; the second call supplies 20 instead. return(result) sends the calculated value back to the caller.

Try a change: predict my_function(2, y = 3) before running it. Expect 5. You changed the arguments in the call; the function definition still has the same default.

Extend a function to handle missing values

Optional extension: the following example adds decisions and missing-value handling to my.mean(). First check that the initial version above and mean() both return 75 for X <- 50:100. If decisions are unfamiliar, revisit if-then-else.

Here na.rm defaults to TRUE. Create X again so this extension has its own complete setup:

X <- 50:100
my.mean <- function(x, na.rm = TRUE){
  if(na.rm == TRUE){
    sum(x, na.rm = TRUE)/length(x[!is.na(x)])
  } else {
    if(any(is.na(x) == TRUE)){
      cat("Error: you need to change your NA settings!")
    }
    
    sum(x)/length(x)
  }
}

X[1:2] <- NA

my.mean(X)
my.mean(X, na.rm = FALSE)

The fixed replacement removes 50 and 51, leaving the values from 52 to 100. With the default, the function divides their sum by the number of non-missing values and returns 76. With na.rm = FALSE, it keeps the missing values, prints the diagnostic sentence and returns NA; the word “Error” in that printed text does not stop R. Compare the two results with mean(X, na.rm = TRUE) and mean(X, na.rm = FALSE).

This is a teaching function for numeric vectors. If all values are missing, removing them leaves no values to average: the calculation gives NaN. Use the documented mean() function in your own analyses.

Test what you’ve learned so far!

Updated: