Taste of R: An Introduction

Date: May 16, 2011

Introduction

R is a statistical programming language and environment, it is open source and available on most platforms. R is not a replacement of Java, C, Perl, Python or other common language; R is a specific tool for data calculations, manipulations and graphing.

If you have a programming background, R can be a great replacement of Excel, it may not be for everyone, but is a great programmable tool powerful for those willing to put the time in. See the R Project site to download and more info.

This article will go through examples on how you can use R to replace Excel. This may not be for everyone, but if you’re like me and love the command-line and vim, you’ll love R.

Quick Examples

# variable assignment
list <- c(1,3,6)  

mean(list)
[1] 3.33    - result 

max(list)
[1] 6       - result

sum(list)
[1] 10      - result

# add each element of the lists together
list2 <- c(2,4,8)
list + list2 
[1]  3  7  14

# create sequence
s <- seq(0,10, by=0.5)
plot(s)

Reading Data

There are numerous ways to get data into R, from reading from text files, to databases and even direct from the internet. Data sets used: pageviews.data, weight.data, movies2010.csv

Read from Text Files

# read in table, single column of data
data <- read.table("pageviews.data")
data[1:5,]    # verify

# plot as timeseries
plot(ts(data))

# read table, specify delimiter
weight <- read.table("weight.data", sep="|", header=TRUE)
plot(weight$Date, weight$Weight)

# read in csv file
data <- read.csv(file="movies2010.csv")
library(ggplot2)
p <- ggplot(data, aes(Box.Office, Rating))
p + geom_point() + 
    scale_x_continuous(
      breaks=c(100000000,200000000,300000000,400000000), 
      labels=c("100M", "200M", "300M","400M"))

Load from Database

library(RMySQL)
con <- dbConnect(dbDriver('MySQL'), 
  user='demo', 
  password = 'demo',
  host = 'localhost',
  dbname = 'baseball')

resultSet <- dbSendQuery(con, 
    "SELECT W,Attendance,yearID,name
     FROM teams
    WHERE (yearID between 1990 and 2010)
      AND franchID = 'SFG' ")

stats <- fetch(resultSet, n=-1)

# plot Wins vs. Attendance
library(ggplot2)
p <- ggplot(stats, aes(x=W, y=Attendance, label=yearID))
p + geom_point() + geom_text(hjust=0.2, vjust=-0.5, size=2.6)

Read from Internet

# grab stock quote data from Yahoo
etfc <- read.csv(paste("http://ichart.finance.yahoo.com/table.csv?", "s=ETFC", "&g=m", sep=""))

# save to file (so dont need to fetch again)
save(etfc, "etfc.RData")

# read in from file
etfc <- load("etfc.RData")

# verify, show first 5 rows
etfc[1:5,]

# simple plot
plot(etfc$Date, etfc$Close)

# better plotting
library("ggplot2")

qplot(as.Date(Date, "%Y-%m-%d"), Close, 
  data=etfc, geom="line", 
  xlab="", ylab="", 
  colour = I("steelblue4"),fill = I("steelblue4"))

Graphing: ggplot

ggplot is a powerful graphing library whose graphs are a bit better looking than R defaults. I tend to use ggplot whenever I can, but will show both methods for this tutorial. Additionally, ggplot has a powerful theme system you can use for consistent colors and styles.

x <- seq(-3, 3, by=0.1)
y <- sin(x)

# normal plot
plot(x,y)

# save a standard plot
png("standard.png")
plot(x,y)
dev.off()

# better plotting
library("ggplot2")
qplot(x,y)
qplot(x,y, geom="line", colour = I("steelblue4"))

# maps
try_require("maps") 
states <- data.frame(map("state", plot=FALSE)[c("x","y")]) 
(usamap <- qplot(x, y, data=states, geom="path")) 

# save with ggplot
ggsave(file="sin.png")

More ggplot2 examples

Bar Chart Side-by-Side

A basic bar chart, the data shown is percentages of time spent per media type compared to advertising dollars spent.

m <- matrix(c(8,27,16,11,43,43,25,19,8,0.5), nrow=2)
colnames(m) <- c("Print","Radio","TV","Internet","Mobile")
barplot(m, beside=T)

If you want two different charts next to each other

library('gridExtra')
plot1 <- ggplot(td, aes(Year, Tablets)) + geom_bar(stat="identity")
plot2 <- ggplot(td, aes(Year, PC)) + geom_bar(stat="identity")
grid.arrange(plot1, plot2, ncol=2)

Plotting Two Lines on Same Graph

An example plotting two lines on to the same graph

# setup data
x <- seq(0, 50, 1)
supply <- x * -2 + 100
demand <- x * 2
df <- data.frame( x = x, supply=supply, demand=demand) 

library(ggplot2)
ggplot(df, aes(x)) + 
  geom_line(aes(y=supply, colour="supply")) + 
  geom_line(aes(y=demand, colour="demand")) +
  opt(title='') +

Cycle Graph

A cycle graph is an interesting way to look at cyclic data such as weekly pageviews of a web site. Typically a web site traffic will see a large dip on weekends which can make it difficult to see what patterns might be occuring.

Here’s an example, the graph on the left is a standard linear graph, on the right is the same day plotted as a cycle graph, the cycle being days of the week. You immediately notice Wednesday dips while the rest are mostly up.

Here’s how the above graphs were created, using pageviews.data

# read in table
data <- read.table("pageviews.data")

# plot as normal timeseries, difficult to see
plot(ts(data))

# cycle plot
monthplot(ts(data, start=1, frequency=7))

Graph: Axis, Labels and Legenes

To add axis labels to the chart, use xlab and ylab, using the stock quote example above with labels

  qplot(as.Date(Date, "%Y-%m-%d"), Close, 
    data=etfc, geom="line", 
    xlab="Year", ylab="Stock Price", 
    colour = I("steelblue4"),fill = I("steelblue4"))

Using ggplot

  ggplot(df, aes(x)) + 
    geom_line(aes(y=supply, colour="supply"), size=2) + 
    geom_line(aes(y=demand, colour="demand"), size=2) +
    scale_x_continuous('') + 
    scale_y_continuous('')

Manipulating Data

# data entry
x <- c(1,2,3)

# more data entry, using stdin (keyboard)
x = scan()  
1:

sort(x)

# diff command
# accumlative total of mail sent
mailings <- c(12345, 23432, 36765, 49567, 60234)
diff(mailings)

# tabulate data
survey <- c("a", "b", "b", "b", "c", "a", "a", "c", "c", "b")
table(survey)

# random numbers
runif(5, 1,10, 5)         # pick 5 numbers between 1 - 10
sample(1:10, 5, replace=F)
sample(1:10, 5, replace=T)

Linear Regression

# setup data
x <- seq(1, 10, by=0.5)

# create random data for Y
y <- seq(10, 46, by=2) + runif(19, 1, 10) 

lm(y ~ x)            # linear model equation
plot(x, y)           # plot data
abline(lm(y ~ x))    # add regression line to graph

# using ggplot
library(ggplot2)
p <- qplot(x, y)
p + stat_smooth(method="lm",size=1)

Programming

# packages (cran)
install.packages("ggplot2")
library("ggplot2")

# conditionals
c <- 42
x <- if (c == 42) 3.14 else 2.71

# loops
teams = c("BAL","BOS","CHW","CLE","DET")
for (team in teams) {
  print(paste("Hello", team, sep=" "))  # string concat
}

Function Example

oddcount <- function(x) { 
  c <- 0 
  for(n in x) {
    if (n %% 2 == 1)
      c <- c + 1
  }
  return(c)
}

set <- c(3,4,5,6,9)
oddcount(set)
[1] 3

Baseball Example

Batting Average Hack from Baseball Hacks by Jospeh Adler

oddcount <- function(x) { 
library(RMySQL)
library(lattice)

con <- dbConnect(dbDriver('MySQL'), 
  user='demo', 
  password = 'demo',
  host = 'localhost',
  dbname = 'baseball')

res <- dbSendQuery(con,
  "select * from batting where yearID=2010 and AB > 250");
batting <- fetch(res, n=-1);
attach(batting);

#Compute batting averages
AVG <- H/AB;

#Plot the charts
histogram(~ AVG | teamID)

densityplot(~ AVG | teamID)

Basketball Example

NBA example, plotting field goal percent and rebounds verse wins.

# load data
data <- read.csv("nba.csv")

# lets look at just 2009
data2009 <- subset(data, year == 2009)  # use conditional

library(ggplot2)

# theory: field goal percent = wins ?
p <- ggplot(data2009, aes(x=o_fgm/o_fga, y=won, label=X.team))
p + geom_point() + geom_text(hjust=0.2, vjust=-0.5, size=2.6)

# theory: rebounds = wins ?
p <- ggplot(data2009, aes(x=o_reb, y=won, label=X.team))
p + geom_point() + geom_text(hjust=0.2, vjust=-0.5, size=2.6)

New User Report Example

Another web site example, graphing new users signing up over last 30 days.
You’ll need your own data source

library(RMySQL)
library(ggplot2)

source("~/.dbconns/prod_slave.R")

# grab data
res <- dbSendQuery(con,
  "SELECT 
     DATE_FORMAT(dt_created, '%m/%d/%Y') as dt, 
     count(*) as c
  FROM users 
  WHERE dt_created > DATE_SUB(NOW(), INTERVAL 30 DAY) 
  GROUP BY dt
  ");
results <- fetch(res, n=-1);

p <- ggplot(results, aes(x=dt, y=c))
p + geom_bar()

Example Examples

Almost every package has built in examples which shows how to use.
Here are just a few, if you ever get stuck check out the examples:

example(plot)
example(abline)
example(pie)
example(spline)

library(ggplot2)
example(qplot)

library(lattice)
example(histogram)

Data Sources

Basketball data from Basketball Database
Baseball data from Baseball Databank

Marcus Kazmierczak