An introduction to R

Functions are R's black box… Take the function mean as example.

mean(iris$Sepal.Length)

[1] 5.843333

Functions are just like other 'commands' in Stata, SPSS or SAS.

SPSS: mean()
Stata: mean; egen mean
SAS: MEAN

• R has evolved so fast that there are thousands of functions.

• Around 250,000 to be more exact! 190 times more than SAS.

• We don't have enough time to cover functions, for that, see here.

• Today we'll cover the basics. Let's start!

Can anyone tell me what does the mean() function do?

• sum() all numbers and divide by the total length() of the vector.
• Create a vector from 1:100 and do it yourself!

mean_vector <- 1:100
sum(mean_vector)/length(mean_vector)

[1] 50.5

How can we turn this into a function?

our_mean <- function(x) {
  sum(x)/length(x)
}

our_mean(mean_vector)

[1] 50.5

mean(mean_vector)

[1] 50.5

Great job!

our_mean <- function(x) {
  sum(x)/length(x)
}

• our_mean is the name of our function
• x is the only argument (but there can be more!)
• Everything inside {} is the code to execute, more formally, the body of the function.

.1 Create a function called adder

.2 It accepts two arguments called x and y

.3 Inside the body, add y and x and don't give with it a name.

adder <- function(x, y) {
  y + x
}

• Generally speaking, what does this function do?

You often create function to avoid repeating code.

Example:

mtcars_two <- mtcars

mtcars_two$cyl <- as.character(mtcars$cyl)
mtcars_two$vs <- as.character(mtcars$vs)
mtcars_two$am <- as.character(mtcars$am)
mtcars_two$gear <- as.character(mtcars$gear)
mtcars_two$carb <- as.character(mtcars$carb)

Transforming, eh? Typical.

• Which things change in this code?

• Write a function called 'to_character'
• It accepts two arguments, old_var and new_var
• The function should contain an expression where you turn a variable into character and save it a new name
• In short, similar to the code from above

First we start with the code that works

old_var <- "cyl"
new_var <- "cyl"

as.character(mtcars$old_var)

Does this work?

as.character(mtcars[, old_var])

Now we have to assign the new name.

mtcars$new_var <- as.character(mtcars[, old_var])

Does this work?

mtcars[new_var] <- as.character(mtcars[, old_var])

Okay, so we got this working…

old_var <- "cyl"
new_var <- "cyl"

mtcars[new_var] <- as.character(mtcars[, old_var])

Wrap it in a function!

to_character <- function(old_var, new_var) {
  mtcars[new_var] <- as.character(mtcars[, old_var])
  mtcars
}

our_mtcars <- to_character(new_var = "cyl", old_var = "cyl") # why did this order change?
class(our_mtcars$cyl)

[1] "character"

All good and well but this only works for the mtcars dataset!

• Add a new argument df to the to_character function
• Replace mtcars with df inside the function

to_character <- function(df, old_var, new_var) {
  df[new_var] <- as.character(df[, old_var])
  df
}

Let's try it with the iris data! This data frame is already available in the working environment. Check head(iris)

our_iris <- to_character(iris, "Species", "Species") # why didn't I name the arguments?
class(our_iris$Species)

[1] "character"

Just as in our own function, functions can have many many arguments or options.

For example..

url <- "https://gist.githubusercontent.com/seankross/a412dfbd88b3db70b74b/raw/5f23f993cd87c283ce766e7ac6b329ee7cc2e1d1/mtcars.csv"
mtcars <- read.csv(file = url, sep = ",", header = TRUE, row.names = 1)

Answer this:

• What's the function name?
• What do each of their arguments do?

When you don't know what a function or its arguments do, search for its help page.

• ?read.csv

Things to consider:

• Read argument definitions
• Checkout the examples
• Run them right away!

• ?mean
• ?sd

With this vector

vec <- sample(c(1:100, NA), 1000, replace = T)

• Calculate the mean and sd (standard deviation)

Help files have several sections you need to be aware of.

• Description *
• Usage *
• Arguments *
• Details
• Value *
• Note
• References
• See also
• Examples *

For example, let's create a data frame. This would be the function to use.

?data.frame

How many arguments have I used?

data.frame(num = 1:10, char = letters[1:10], sample(c(T, F), 10, replace = T))

What changed from the example in the help document?

data.frame(num = 1:10, char = letters[1:10], sample(c(T, F), 10, replace = T),
           row.names = 1, check.rows = TRUE, fix.empty.names = FALSE)

In the RECSM seminars you'll be using some advanced R which is why we need to take you to the limit!

• Run one example with the lm (Fitting linear models) function and the mtcars dataset.
• Use by to split mtcars by the factor cyl and apply the summary function
• Create a new variable in mtcars called mpg_mean using ifelse. It gives back a 1 when mpg is above or equal to the mean and 0 when it's not.

Remember to use ?function

lm(mpg ~ vs + cyl, data = mtcars)
by(mtcars, mtcars$cyl, summary)
mtcars$mpg_mean <- ifelse(mtcars$mpg >= mean(mtcars$mpg), 1, 0)

Packages are one of the most important things in R.

• They allow people to share ideas/code
• They are well documented
• They can contain functions or datasets

Where are R packages? In something called CRAN (Comprehensive R Archive Network)

How do you install them?

install.packages("cowsay")
install.packages("lme4")

How do you use them? Once installed we will have to call them in order to get them running in the current session.

library("cowsay")
library("lme4")

Here we have some more info provided by the help documents.

?cowsay::say
?lme4::nlmer

Read a bit, and then check the examples!

How do we repeat things?

for (column in mtcars) {
  if (is.numeric(column)) {
    print(is.numeric(column))
  } else {
    message("Not numeric")
  }
}

Let's explain it in the console…

I think you're ready for some real R programming…