Live code:

Creating functions

Creating functions in R

Here we will learn how to write functions in R. Functions are extremely helpful for automating commons tasks that you might use often (e.g. computing the RMSE for a set of predictions). If you’re going to use the same block of code more than twice, you should consider writing a function!

There are three key steps to creating a new function:

1. Picking a NAME for the function
2. Listing the INPUTS/ARGUMENTS to the function called function()
3. Placing the code you have developed in the BODY of the function (between the sets of curly braces { } that immediately follow function()
   1. Making sure to return() the output from the function

Example

For example, suppose that I want to create a function that takes in a matrix and returns the sums of the values within each column. (There is a function that already does this, but pretend there isn’t!) The function might look like this:

column_sums <- function(input_matrix){
  n_cols <- ncol(input_matrix)
  sums <- rep(NA, n_cols)
  for(i in 1:n_cols){
    sums[i] <- sum(input_matrix[,i])
  }
  return(sums)
}

This looks scary, but let’s break it down.

• The name of the function is column_sums.

• R knows I want to create a function because I use the function() function, where I specify that my function requires a single input that will be referred to as input_matrix.

• The code within the body specifies how I will use input_matrix to calculate the column sums.

• I finish the function by return()-ing a vector.

Let’s test this out: I will create a matrix of numbers, and then use/call my function as I normally would any R function:

my_mat <- matrix(1:10, nrow = 2, ncol = 5)
my_mat

     [,1] [,2] [,3] [,4] [,5]
[1,]    1    3    5    7    9
[2,]    2    4    6    8   10

column_sums(my_mat)

[1]  3  7 11 15 19

Let’s confirm this is the correct output with the pre-provided R function colSums():

colSums(my_mat)

[1]  3  7 11 15 19

How might we code a function that calculates the squared error between two values? Try it yourself, then check here to see if your code generally agrees with mine:

Code

squared_error <- function(x1, x2){
  ret <- (x1 - x2)^2
  return(ret)
}

Code

# test your code: you should get 16 for passing in -2 and 2
squared_error(-2, 2)

Your turn!

Feel free to try any and all of the following:

1. Write a function that takes in a temperature in degrees Fahrenheit and returns the temperature in degrees Celsius. For reference, the conversion is \((\text{degreesF} - 32) * \frac{5}{9}\). Give your function an appropriate name.
   1. Check to make sure your function works by passing in \(32^\circ\) F. You should get 0 back!
2. Write a more complicated version of Exercise 1 where the function takes in two inputs: 1) a temperature (in either Fahrenheit or Celsius) and 2) a string or Boolean (your choice!) that denotes if you want to convert to Fahrenheit or Celsius. Your function should return the correct conversion based on the user’s inputs. Note: the conversion from Celsius to Fahrenheit is \(\text{degreesC} * \frac{9}{5} + 32\).
3. Writ a function called get_rmse() that takes in two vectors as inputs: one of predictions, and one of true values. Your function should calculate and return the root mean squared error (see slides for equation).
4. Last week we practiced coding for() loops by obtaining the factorial of a given positive integer. Create a function where the user specifies the integer they want to find the factorial of, and return the factorial.