R for Statistical Computing

HMC Scientific Computing Workshop Series, Fall 2014

(Slides are made with the R Presentations tool in RStudio) (Some materials are adapted from the R-Bootcamp by Jared Knowles: http://jaredknowles.com/r-bootcamp/)

• Slides at http://bit.ly/hmc-r-workshop-slides
• Login at http://bit.ly/hmc-r-workshop-sign-in
  • Type your name and email and click Next
• Files at http://bit.ly/hmc-r-workshop-files
• Homework assignment submission at
  • http://bit.ly/hmc-r-workshop-homework
• Digital badge requirements
  • Attend the workshop (login)
  • Work (R file submission)

• What is R?
• What is not R?
• Then Why R?
• What is RStudio?
• What can you do with it?
• I want to install it!
• Look Ma, R can do MATH!
• Even more Math!

• R is a statical programming language/environment.
• R is open source/free.
• R is widely used/prefered.
• R is cross-platform.
• R is hard to learn (really?).

• S: R's ancestor
• S-Plus: Commercial; modern implementation of S
• SAS: Commercial; widely used in the commercial analytics.
  
• SPSS: Commercial; easy to use; widely used in Social Science.
• MATLAB: Commercial; can do some Stats.
• Python: Also can do some Stats; good in text data manipulation.

• R-help listserv (Google r-help)
  
• Stack Overflow: http://stackoverflow.com/questions/tagged/r
• Cross-Validated: the statistics Q&A site http://stats.stackexchange.com/
• HMC Scientici Computing Specialist: jepark@hmc.edu

• Integrated Development Environment for R
• Nice combination of GUI and CLI
• Free and commercial version
• 4 main windows, tabs, etc
• Version control: Git and VPN
• R Markdown
• R Presentation

• R for Mac
• R for Linux
• R for Windows
• RStudio Desktop
• RStudio Server

Google is our friend!

• packages are add on features to R that include data, new functions and methods, and extended capabilities. Think of them as “apps'' on your phone. We've already installed several!
• terminal is the main window of R where you enter commands
• scripts are where you store commands to be run in the terminal later, like syntax files in SPSS or .do files in Stata
• functions are commands that do something to an object in R
• dataframe is the main element for statistical purposes, an object with rows and columns that includes numbers, factors, and other data types

• workspace is the working memory of R where all objects are stored
• vector is the basic unit of data in R
• symbols is used to name and store objects or to designate operations/functions
• attributes determine how functions act on objects

1+1

2+runif(1,0,1)

2+runif(1,min=0,max=1)

3^2

3*3

sqrt(3*3) # comments

# comments are preceded by hash sign

• R can take integrals and derivatives, for example:

Numerical Integral of

\( \displaystyle\int_0^{\infty} \frac{1}{(x+1)\sqrt{x}}dx \)

integrand <- function(x) {1/((x+1)*sqrt(x))} ## define the integrated function

integrate(integrand, lower=0, upper=Inf) ## integrate the function from 0 to infinity

3.142 with absolute error < 2.7e-05

demo() # display available demos

demo(graphics) # try graphics demo

library() # show available packages on the computer

search() # show loaded packages

?hist # search for the usage of hist function

??histogram # search for package documents containing the word "histogram"

R workspace stores objects like vecors, datasets and functions in memory (the available space for calculation is limited to the size of the RAM).

a <- 5 # notice a in your Environment window

A <- "text" 

a
A
ls()
print(c(a,A))
print(a,A)

VECTOR (homogeneous)
A vector is an array object of the same type data elements.

class(a)

class(A)

B <- c(a,A) # concatenation

print(B)

class(B) # why?

LIST (heterogeneous)
A list is an object that can store different types of vectors.

aList <- list(name=c("Joseph"), married=T, kids=2)
aList
aList$kids <- aList$kids+1
aList$kids
aList <- list(numeric_data=a,character_data=A)
aList

Data Frame
A data frame is used for storing data tables. It is a list of vectors of equal length.

n <- c(2, 3, 5) # a vector 
s <- c("aa", "bb", "cc") # a vector
b <- c(TRUE, FALSE, TRUE) # a vector
df <- data.frame(n, s, b) # a data frame
df
mtcars # a built-in (attached) data frame
mtcars$mpg

Data Frame (cont.)

myFrame <- data.frame(y1=rnorm(100),y2=rnorm(100), y3=rnorm(100))
head(myFrame) # display first few lines of data
names(myFrame) # display column names
summary(myFrame) # output depends on the data types
plot(myFrame)
myFrame2 <- read.table(file="http://scicomp.hmc.edu/data/R/Rtest.txt", header=T, sep=",")
myFrame2

FACTOR

• Factors are a special compoud object used to represent categorical data such as gender, social class, etc.
• Factors have 'levels' attribute. They may be nominal or ordered.

v <- c("a","b","c","c","b")
x <- factor(v) # turn the character vector into a factor object
z <- factor(v, ordered = TRUE) # ordered factor
x
z
table(x)

Single Sample Tests

Our questions for the sample might be:

• What is the average (mean)?
• Is the mean value significanntly different from current expectation?
• What is the level of uncertainty associated with our estimate of the mean value?

Single Sample Tests

For our valid inferences, we need to find some facts about the distribution of data:

• Are the values normally distributed or not? ()
• Are there outliers in the data? (Should we safely exclude them?)
• If data were collected over a period of time, is there evidence for serial correlation? (Do we need to do time series analysis?)

Based on the facts, standard parametric tests like Student's t test might not be applicable and we may have to seek for non-parametric techniques.

Single Sample Tests

data <- read.table(file="http://scicomp.hmc.edu/data/R/das.txt", header=T)
names(data)
attach(data)
par(mfrow=c(2,2))
plot(y)
boxplot(y)
hist(y,main="")
y2 <- y
y2[52] <- 21.75
plot(y2)
dev.off() # reset the graphic device

Single Sample Tests

summary(y)
# Graphical test for normality
qqnorm(y)
qqline(y,lty=2)
# Test for normality
x <- exp(rnorm(30)) # lognormally distributed 
shapiro.test(x) # look at the p value

The null-hypothesis of this test is that the population is normally distributed. Thus if the p-value is less than the chosen alpha level (e.g., p < 0.05), then the null hypothesis is rejected and there is evidence that the data tested are not from a normally distributed population.

Please submit your R environemnt file at http://bit.ly/hmc-r-workshop-homework (a digital badge requirement).

The file has .RData extension. To attach to the form, it must be changed to .txt.

Useful links:

• R tutorial site: http://jaredknowles.com/r-bootcamp/ and
• R search site: http://rseek.org
• R cheat sheets: http://devcheatsheet.com/tag/r/
• R Markdown cheat sheet: http://shiny.rstudio.com/articles/rm-cheatsheet.html