• http://www.ats.ucla.edu/stat/r/library/intro_function.htm
• http://adv-r.had.co.nz/Functions.html
• http://www.statmethods.net/management/userfunctions.html
• http://www.stat.berkeley.edu/~statcur/Workshop2/Presentations/functions.pdf
• http://www.r-bloggers.com/how-to-write-and-debug-an-r-function/
• http://cran.r-project.org/doc/contrib/Lemon-kickstart/kr_rfunc.html
• http://nicercode.github.io/guides/functions/

• What is a function?
• Basic examples
• Returning values
• Local and global values
• Debugging
• Location and organization
• Source()

• Basic structure

name.of.function <- function(argument1, argument2) {
  statements
  return(something)
}

• Make the name informative
• Arguments can be single values, vectors, matrices, data frames, lists
• Statements are what you want done
• It can be just a single line of code or many
• Try not to make it too long
• Don't forget the return() at the end!

• Make a simple function

fun1 <- function(x) {
x*2  
3 * x - 1
}
fun1(5)

## [1] 14

• Only the last expression evaluated is returned
• Note that we didn't save the calculation to an object
• So it gets automatically returned to the console

• Same function except we saved it to an object

fun2 <- function(x) {
  y <- 3 * x - 1
}
fun2(5)

• Doesn't return anything because the object was a local variable
• Local variables are defined inside {}
• E.g. For loops and functions

• Two ways to resolve this "issue"
• 1. Recommend using return()

fun2.1 <- function(x) {
a <- x*2
b <- paste(a, "is my favorite number")
y <- 3 * x - 1
return(y)
}
fun2.1(5)

## [1] 14

• But a and b were cannot be returned…

fun2.1 <- function(x) {
a <- x*2
b <- paste(a, "is my favorite number")
y <- 3 * x - 1
return(list(a,b,y))
}
fun2.1(5)

## [[1]]
## [1] 10
## 
## [[2]]
## [1] "10 is my favorite number"
## 
## [[3]]
## [1] 14

• Can also save it to a global object

fun2.2 <- function(x) {
Y <<- 3 * x - 1
}
fun2.2(6)

• Use <<-
• Convention is to use CAPITAL names like Y
• Not recommended - Only seen this occasionally used in for loops

• Goal is to create a function that welcomes you to R
• There are some functions built into R
• http://www.statmethods.net/management/functions.html

#Gives your system's information and the R version you are using
myinfo <- Sys.info() 
myinfo[7]

##        user 
## "Nick.Sard"

rinfo <- R.Version()
rinfo$nickname

## [1] "Sock it to Me"

hiyou <- function(){
  #extracting some information
  myinfo <- Sys.info() ; rinfo <- R.Version()
  
  #making three statements
  hi <- paste("Hi",paste0(myinfo[7],"!"))
  welcome <- paste("Welcome to",paste0(rinfo$version.string,"!"))
  iam <- paste("My name is:",paste0(rinfo$nickname,"."),
               "Let's go have some fun!")
  
  #returning a cat statement
  return(cat(paste0(hi,"\n",welcome,"\n",iam)))
}
hiyou()

## Hi Nick.Sard!
## Welcome to R version 3.1.1 (2014-07-10)!
## My name is: Sock it to Me. Let's go have some fun!

mysummary <- function(x, output=1) {
  #calculating stats of interest
  center <- mean(x) 
  spread <- sd(x)
  
  #Using output to determine how the information is
  #displayed to the user
  if (output == 1) {
    return(cat("Mean=", center, "\n", "SD=", spread, "\n"))
  } 
  
  if (output == 2) {
    df <- data.frame(statistic = c("mean","sd"),
                     value = c(center,spread))  
    return(df)
  }
}

set.seed(1234)
x <- rnorm(n = 500,mean =  4,sd = 1) 
mysummary(x)

## Mean= 4.001839 
##  SD= 1.034814

mysummary(x, output=2)

##   statistic    value
## 1      mean 4.001839
## 2        sd 1.034814

mysummary(output=2,x)

##   statistic    value
## 1      mean 4.001839
## 2        sd 1.034814

• We can add some if statements to prevent errors

mysummary <- function(x, output=1) {
  #calculating stats of interest
  center <- mean(x) 
  spread <- sd(x)
  #Using output to determine how the information is
  #displayed to the user
  if (output == 1) {
    return(cat("Mean=", center, "\n", "SD=", spread, "\n"))
  } 
  if (output == 2) {
    df <- data.frame(statistic = c("mean","sd"),
                     value = c(center,spread))  
    return(df)
  }
}

• Use if() to make sure the right type of data is being used. When it isn't -
• stop() - stops function
• warning() - adds a warning
• print() can provide feedback to user

mysummary <- function(x, output=1) {
if(is.numeric(x)!=T){
stop("x is not numeric")
}
if(length(x)<30) {
warning("Sample size is <30",call. = F)
} 
print("Calculating mean and standard deviation")
# the rest of the function...
}

mysummary(x = rnorm(100,4,1))

## [1] "Calculating mean and standard deviation"
## Mean= 3.863123 
##  SD= 0.9289288

#mysummary(x = LETTERS[1:10])
mysummary(x = c(1,2,3))

## Warning: Sample size is less than 30

## [1] "Calculating mean and standard deviation"
## Mean= 2 
##  SD= 1

• Goal calculate geometric mean across all columns

gm_mean = function(x, na.rm=TRUE){
exp(sum(log(x[x > 0]), na.rm=na.rm) / length(x))
}
# column geometric means - apply() takes matrices
apply(as.matrix(mtcars[,1:4]), MARGIN = 2, gm_mean)

##        mpg        cyl       disp         hp 
##  19.250064   5.919439 197.321758 131.883679

#compared to mean
apply(as.matrix(mtcars[,1:4]), MARGIN = 2, mean)

##       mpg       cyl      disp        hp 
##  20.09062   6.18750 230.72188 146.68750

• Fibonacci sequence example

fib <- function(n){
#base case
if(n==1){
return(1)
} else if(n==2) {
return(1)
#inductive case
} else {
a = fib(n-1)
b = fib(n-2)
return(a+b)
}
}
c(fib(1),fib(2),fib(3),fib(4),fib(5),fib(6))

## [1] 1 1 2 3 5 8

• Keep functions short
  • Split up function into easy to understand chunks
  • Can call other functions
  • Code is cleaner and easy to test and update
• Put in comments so that user knows:
  • What the inputs to the function are
  • What the function does
  • What the output(s) are
• Check for errors
  • Test function with simple examples
  • Use debugging and error messages

• Throughout R script
• Recommended at the beginning
• Single or multiple function(s)
• Save in another script just for functions
• Single or multiple function(s)

• Make an R script containing all your functions
• Call at the beginning of your scripts

source("my.functions.R")

### function.name()

#Description: This function does some really cool stuff! Details...

#Inputs:
# df - a data frame with three columns
#   column 1 - population same
#   column 2 - sample ID
#   column 3 - length

#Outputs
# options - Default is 1
#   option 1 - returns mean and standard deviation for length
#   option 2 - creates a histogram of lengths