Writing functions in R

Some handy advice for creating your own functions

A quick review of R

R is a statistical software package and an object-oriented programming language

Terms to remember: Vectors, matrices, and dataframes Indices Functions

Warm up

Download the data for lab 3 In Rstudio, go to Workspace → Import Dataset

→ From Text File Make sure to select the header option If you're not using Rstudio, the code is:

data_lab_3 <- read.csv("~/documents/classes/Psych 1950/mood.csv")

Where ~ is the path name

Warming up a little more

Use the help() function to read about the read.csv() function

How could we use it to read in a file with no header?

read.csv(“filename”,header=FALSE) We can also use R to read in SPSS files, but

for now we'll stick with read.csv()

Last page of warm-up (I promise!)

Find the standard deviation (sd()) of the second column (puDay2call1) of your dataframe

Uh-oh! That output isn't helpful Add the following argument to the standard

deviation function: na.rm=TRUE

A slight modification

Suppose that we want to calculate the standard deviation using the population formula

Check the help file for sd(). Is there a way to do that?

Nope! We'll need to make our own....

Making a function

Let's start with something easier We'll make our own mean() function What should it do?

We'll pass* it a vector of numbers as arguments*

It should return* the mean

*programming jargon

The function syntax

getMean <- function(arguments){

commands go here

} The name of the function is getMean() (this

is usually a verb) The arguments are the values and instructions

we give to the function The body is where the work happens

Iteration 1

getMean <- function(x){

return(sum(x)/length(x))

} Try this on the second column How can we handle NAs in the function,

assuming we ALWAYS want to remove them?

Iteration 2

getMean <- function(x){

return(sum(x,na.rm=T)/length(x))

} Now try this one, and compare your results to

R's built-in mean function Why aren't the values the same?

Hint: what's the length of a vector that contains NAs?

Iteration 3

getMean <- function(x){

return(sum(x,na.rm=T)/length(na.omit(x))

} Another R function saves the day! Thanks, R! Compare your results to the built-in function

Another way to do it

We've been leaning heavily on the sum() function

Sometimes, though, we need to tell R to do a certain operation a number of times

To do that, we use an operation called a for loop

There are other loops as well, but we'll stick with a for loop

The anatomy of a for loop

getFactorial <- function(number){

j=1

for (index in 1:number){ j <- j*index

}

return(j)

} What will this function do?

One more concept

Sometimes, we need a function to make a decision

Here, we use conditionalsif(condition){ #if the condition is true

Something #do this

}

else{ #if it's false

something else #do this instead

}

For examples

if (!is.na(x)){ #if x isn't an NA

print(x) #write x. If it is, nothing

} #will happen

if (x<=4){ #if x is less than 4

print(x-1)

}

if (x==5){ #if x is exactly 5

print(“Five”)

}

Looping to get the mean

getMean_3 <- function(x){

sum <- 0

length <- 0

for (i in 1:length(x)){

if (!is.na(x[i])){ #exclude NAs

sum <- sum+x[i] #keep a running tally of the sum

length <- length+1 #and the length

}

}

return(sum/length) #this is the mean

}

Adding some complexity

It's your turn now: Write two functions to compute the sum of

squared deviations from the mean of a vector In one version, use the sum() function In the other, use a for loop

Try to allow your function to work with a vector that includes some NAs

Remember

The formula for the sum of squares of a set of numbers is the sum of (x

i – mean(x))2

Now make R do it for you!

Last of all

Make a new function that finds the (population) standard deviation of the vector

Find the sum of squares, divide by the number of observations, and take the square root

Test your function to make sure it's working