Practice Set: Writing Custom Functions

Writing Functions

By now we’re very familiar with how to use functions from base R and installed packages. Before we start creating our own functions, let’s review what functions are composed of:

• Function name: The name of the function, which is used to call the function after it is defined (for example, ‘head’ is the name of the head() function).
• Arguments: Arguments are placeholders. If arguments are a defined part of a function, then values (typically) must be assigned to them. Arguments can also have default values, in which case values may not need to be designated.
• Function body: The body of a function is a collection of statements that determine what the function does.
• Return value: The return value of a function is the last expression in the function body to be evaluated. It is whatever is output by the function.

The syntax for defining a function is as follows:

function_name <- function(arg_1, arg_2, ..., arg_x) {
  Function body 
  return(your_output)
}

For example, maybe we ran a study with 50 subjects, and subjects with odd number IDs were randomly assigned to Counterbalance A and subjects with even number IDs were randomly assigned to Counterbalance B. We forgot to include this information in our data.frame (called studyData), so now it needs to be added. There are 50 participants total (50 rows), and the first few rows look like this:

##   subject    score
## 1       1 72.58587
## 2       2 75.55486
## 3       3 77.16888
## 4       4 70.30860
## 5       5 75.85825
## 6       6 76.01211

Put simply: We want to create a function that determines whether numbers in a vector are even or odd. The simplest version of this function (below) requires no arguments, and will return TRUE for subjects with even numbers and FALSE for subjects with odd numbers.

# Creating the even_odd function
even_odd <- function() {
   return((studyData$subject %% 2) == 0)
}

# Using the even_odd function
even_odd()

##  [1] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [13] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [25] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [37] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [49] FALSE  TRUE

What are the components of this function?

• even_odd is the name
• No arguments are defined, thus the empty parentheses
• (studyData$subject %% 2) == 0 is the body of the function and is nested in…
• return is your return statement

Nothing runs if you only execute the first 3 lines! You need to then call your function if you want it to actually run.

This function may be useful to someone else who regularly needs to check whether their subject numbers are even or odd… but this seems unlikely. Importantly, the function assumes that the name of the data.frame will always be studyData, and that the name of the subject column will be subject. The reason for this assumption? Because studyData and subject are hard-coded into the body of the function!

Wouldn’t it be better if the user could supply their own vector of numbers to the function? We can improve this function by specifying an argument, which we will just call x for now.

# Improving the even_odd function
even_odd <- function(x) {
   return((x %% 2) == 0)
}

# Using the even_odd function for studyData
even_odd(x = studyData$subject)

##  [1] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [13] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [25] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [37] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE
## [49] FALSE  TRUE

# OR, using the even_odd function for another vector of numbers
my_numbers <- 103:109
even_odd(x = my_numbers)

## [1] FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE

How else can we improve upon the current function?

Perhaps we want to return the strings "counterbalanceA" and "counterbalanceB", instead of TRUE and FALSE. Now, we need to incorporate a for loop and an if/else statement into the body of our function. (If you don’t remember the syntax for if/else statements, revisit the practice set for for loops!)

# Improving the even_odd function
even_odd <- function(x) {
  new_vector <- c() # initialize the vector
   for(i in x) {
     if((i %% 2) == 0){
       new_vector <- append(new_vector, "counterbalanceA")
     } else {
       new_vector <- append(new_vector, "counterbalanceB")
     }
   } 
  return(new_vector)
}

# Let's try out our improved even_odd function, and we'll store it as a NEW column called 'counterbalance' to studyData
studyData$counterbalance <- even_odd(x = studyData$subject)
head(studyData)

##   subject    score  counterbalance
## 1       1 72.58587 counterbalanceB
## 2       2 75.55486 counterbalanceA
## 3       3 77.16888 counterbalanceB
## 4       4 70.30860 counterbalanceA
## 5       5 75.85825 counterbalanceB
## 6       6 76.01211 counterbalanceA

The new version of this function is much more generalizable. One additional improvement we could make is to add more arguments that allow the user to choose what the labels for even and odd numbers are. Additionally, we should make sure that our arguments are clearly named!

# Improving the even_odd function (one more time!)
even_odd <- function(vectorName, evenLabel, oddLabel) {
  new_vector <- c() # initialize the vector
   for(i in vectorName) {
     if((i %% 2) == 0){
       new_vector <- append(new_vector, evenLabel)
     } else {
       new_vector <- append(new_vector, oddLabel)
     }
   } 
  return(new_vector)
}

# Make new group column in studyData
# For example, maybe subjects were randomly assigned to a "training" versus a "control" group
studyData$group <- even_odd(vectorName = studyData$subject,
                            evenLabel = "training",
                            oddLabel = "control")

# View data.frame
head(studyData)

##   subject    score  counterbalance    group
## 1       1 72.58587 counterbalanceB  control
## 2       2 75.55486 counterbalanceA training
## 3       3 77.16888 counterbalanceB  control
## 4       4 70.30860 counterbalanceA training
## 5       5 75.85825 counterbalanceB  control
## 6       6 76.01211 counterbalanceA training

head(iris)

##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
## 5          5.0         3.6          1.4         0.2  setosa
## 6          5.4         3.9          1.7         0.4  setosa

# Changing theme_classic into theme_drew
theme_drew <- function(){
  require(ggplot2)
  require(extrafont)
  loadfonts(device = "postscript", quiet = TRUE)
  theme_classic(base_size = 14, base_family = "Georgia")
}

# Below creates 2 plots -- one using the original `theme_classic()` and
# the other using the new custom `theme_drew()`. Then we'll combine them into
# a single figure! 

plot1 <- ggplot(studyData, aes(x = score)) +
  geom_histogram(color = "black", fill = "plum4") +
  theme_classic() +
  ylim(c(0,6)) +
  ylab("Count") +
  xlab("Value") +
  ggtitle("Classic theme")


plot2 <- ggplot(studyData, aes(x = score)) +
  geom_histogram(color = "black", fill = "plum4") +
  theme_drew() +
  ylim(c(0,6)) +
  ylab("Count") +
  xlab("Value") +
  ggtitle("Drew's custom theme")

ggarrange(plot1, plot2)

theme_drew <- function(){
  theme_classic(base_size = 14, base_family = "Georgia")  
        theme(
            # change stuff here
        )
}

# Changing theme_classic into theme_drew
theme_shelly <- function(){
  theme_classic()
        theme(
            panel.background  = element_rect(fill = "transparent", color = NA),  
            plot.background = element_rect(fill = "transparent", color = NA),
            text = element_text(family = "Times New Roman")
        )
}

# Let's give it a go:
plot1 <- ggplot(studyData, aes(x = score)) +
  geom_histogram(color = "black", fill = "plum4") +
  theme_classic() +
  ylim(c(0,6)) +
  ylab("Count") +
  xlab("Value") +
  ggtitle("Classic theme")

plot2 <- ggplot(studyData, aes(x = score)) +
  geom_histogram(color = "black", fill = "plum4") +
  theme_shelly() +
  ylim(c(0,6)) +
  ylab("Count") +
  xlab("Value") +
  ggtitle("Shelly's custom theme")

ggarrange(plot1, plot2)