FunctionsinR

Functions in R

Based off of this tutorial https://nicercode.github.io/guides/functions/

Example of code that does not use functions https://nicercode.github.io/guides/functions/before.html.

Create a basic function

sum.of.squares <- function(x,y) {
  x^2 + y^2
}

sum.of.squares(-3,4)

## [1] 25

Procedure for writing functions:

1. Define the function
2. Load the function into R
3. Use the function

Define a function

function.name <- function(arg1, arg2, arg3=2, ...) {
  newVar <- sin(arg1) + sin(arg2) # do some stuff
  # return(newVar/arg3) # return value using return()
  newVar/arg3 # return value not using return()
}

function.name(1,2)

## [1] 0.8753842

Import an example function from another file

source('examplefunction.R')
exp_absolute_diff(10,15)

## [1] 148.4132

The Ellipsis argument

red.plot <- function(x,y,...) {
  plot(x,y,col='red',...)
}

# red.plot with only specified arguments
red.plot(1:10,1:10)

# red.plot making use of the ellipse
red.plot(1:10,1:10,xlab='My X Axis',ylab='My Y Axis')

Add comments to code to increase readability

exp_sin_plot <- function(x,y,color){
  # PURPOSE:    This function takes a vector x and plots the function exp(y*sin(x))
  # 
  # INPUT:      x, a vector
  #             y, a scalar
  # OUTPUT:     a plot of exp(y*sin(x))

  f_x <- exp(y*sin(x))
  plot(x,f_x,xlab='X',ylab=paste('exp(',y,'*sin(X))'),type='o')
}

exp_sin_plot(seq(1,20,length=200),2)

Functions help avoid coding errors

p <- .2

logit <- function(p){
  log(p) - log(1-p)
}

logit(.3)

## [1] -0.8472979

Loops in R

Example code calculating the square of even numbers in a sequence

x <- seq(1,1000)
even_x_square <- c()

# Use a for loop
start_time = Sys.time()

for(i in x) {
  
  if(i %% 2 == 0){
    even_x_square <- append(even_x_square,i**2)
  }
  
}

end_time = Sys.time()

print(end_time-start_time)

## Time difference of 0.005947828 secs

In general when working with large datasets you DO NOT want to use loops. Vectorizing code is much quicker.

Vector Operations

start_time = Sys.time()

even_x_square <- x[(x%%2)==0]**2

end_time = Sys.time()

print(end_time-start_time)

## Time difference of 0.001984119 secs

Automate analysis with functions and loops.

# Stock market data from last class

Smarket = read.csv(file='Smarket.csv')

run_smarket_analysis <- function(dataset,x_variables){
  
  regression_formula <- as.formula(
                  paste('Direction', 
                  paste(x_variables, collapse = " + "), 
                  sep = " ~ "))
  
  train=(dataset$Year<2005)
  dataset.2005 = dataset[!train,]
  Direction.2005 = dataset$Direction[!train]
  
  glm.fits = glm(regression_formula,
                 data=dataset,
                 family=binomial,
                 subset = train)
  
  glm.probs = predict(glm.fits,dataset.2005,type="response")
  glm.pred=rep("Down",252)
  glm.pred[glm.probs>.5]="Up"
  mean(glm.pred==Direction.2005)
  
}

variable_groups <- list(c('Lag1'), 
                        c('Lag2'),
                        c('Lag1','Lag2'),
                        c('Lag1','Lag2','Lag3','Volume'))

for(vars in variable_groups){
  cat("\n\n")
  print(vars)
  cat('Test Accuracy',run_smarket_analysis(Smarket,vars))

}

## 
## 
## [1] "Lag1"
## Test Accuracy 0.5396825
## 
## [1] "Lag2"
## Test Accuracy 0.5912698
## 
## [1] "Lag1" "Lag2"
## Test Accuracy 0.5595238
## 
## [1] "Lag1"   "Lag2"   "Lag3"   "Volume"
## Test Accuracy 0.4761905