ODSC Workshop: Intro to R

Paul Paczuski [AMA!]

Quick Start

This is a hands-on introduction to R. We will go through actual code and learn about data manipulation, graphics, and fundamentals of statistical modeling.

Set-up/Installation instructions are below.

A sample directory structure is in /templates

Contents

• Set-up
• Installing the R environment
• Outline/Programs
• Resources
• Credits

Set-up

1. Install R, RStudio, and the dplyr and ggplot2 packages [see Installing the R environment below]

2. Download this repository ("Download Zip" link on the top right of this page) and unzip it to a convenient location on your computer. We will be working with its contents.

3. Open R Studio

4. Set the working directory to the /programs directory of the just-downloaded repository, as follows:

   • Session -> Set Working Directory -> Choose Directory...

   • Navigate and Open the /programs directory

5. Open the file 0-intro.R [File - Open file...]. You will find it in /programs

Installing the R environment

Below are instructions to install R, a few packages, as well as RStudio, which we will be using during the workshop.

1. Install R

   Mac OS X
   download and install from:
   http://cran.r-project.org/bin/macosx/ [click R-3.2.0.pkg on left]

   Windows
   download and install from:
   http://cran.r-project.org/bin/windows/base/

   Linux, etc
   See top of:
   http://cran.us.r-project.org

2. Install the following R packages (see instructions which follow):

   ggplot2

   dplyr

   a) MAC OS X and Linux:

   • Open R (not RStudio for this step)
   • In Menu, go to “Packages and Data” - > “Package Installer”
   • Search for and install the above two packages (may need to choose a “mirror” - click on something in the USA) as follows:
   • Type in the name of one package, click “get list”, check “Install Dependencies” and then “Install Selected”
   • Do the same for the other package

   b) Windows:

   • Open R (not RStudio for this step)
   • In Menu, go to "Packages" -> "Install package(s)..." and select each of the packages at top to install.

3. Install RStudio (a great interface for using R)

   Install for your appropriate system from the list at: http://www.rstudio.com/products/rstudio/download/

Outline/Programs

There are four R files in /programs. They will function as an "interactive notebook" for this workshop. Details follow.

• 0-intro.R

• 1-data.R

• 2-graphics.R

• 3-stats.R

0-intro.R

This will be a hands-on Introduction to R

• "R is a free software environment for statistical computing and graphics"
• "The best way to learn a new language is to try out the commands"

3 topics covered:

1. data manipulation, including package dplyr
2. graphics, including package ggplot2
3. basic statistical models: linear and logistic regression

Go through code in the program

References:

• James et al, p 42+ [see Resources section below]

1-data.R

Introduction to data manipulation

• using base R functions

• using intuitive, fast methods from dplyr

References:

• James et al, p. 47

• > vignette("introduction", "dplyr")

• Data wrangling handout

2-graphics.R

Hands-on introductions to creating graphics in R

• using base R methods

• using the very popular ggplot2 package

References:

• James et al, p 45, 49

• http://docs.ggplot2.org/current/

3-stats.R

We use statistical analysis for:

• inference - making conclusions based on data

• prediction - what will happen when I observe new data?

...and we create models to do both of those things.

"All models are wrong - some are useful."

Fundamentals:

• model selection - which model is good/best?

• model diagnostics/validation - is my current model reasonable and does it work?

• uncertainty is always part of the final product

References:

• James et al, p 15 [statistical modeling]

• James et al, p 59+ [Linear regression]

• James et al, p 130+ [Logistic regression]

Resources

Data Wrangling handout
http://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf

Quick-R
http://www.statmethods.net

An Introduction to Statistical Learning with Applications in R, by James et al:
http://www-bcf.usc.edu/~gareth/ISL/

Credits

Matthew Eaton for help with some workshop content
Hadley Wickham: dplyr, ggplot2, etc etc
ISLR book authors