R is a language that's easy to learn badly

Why R?

The old way...

Why R?

A better way

glm(y ~ -1 + a + c + z + a:z, data = mydata, maxit = 30)

This is reproducible, repeatable and will make sense to you (and everyone else) further down the line.

Establish a workflow

A workflow (best suited to you) will help keep your analysis streamlined and reproducible.

A sample workflow

load.r
	
load all raw untransformed data.

clean.r

Transform and clean messy data incl. missing values.

functions.r

Keep all functions in one separate file.

do.r

Your actual analysis. Load cleaned data and source all functions.

# File: do.r
source("functions.r")
load("cleaned_data.rdata")

Avoid restoring workspaces

Workflow...

Set working directores from the console and not scripts

# My_script.r
setwd("path/to/dir")
# Removing any extra objects from my workspace (just in case)
rm(list = ls())
# Keeps my current file clutter free
source("related_r_scripts.r")

Workflow...

Never attach data

attach(mydata)

Always refer to it explicitly

mydata$column_name

Environments and Namespaces

Global Environment

			
a <-  rnorm(100) 

Function or Namespace

	
a <- 5

Annotate your code clearly

Good ✔
# Script to analyze rainfall data.
# loading lib: plyr_1.7.1,  devtools_0.6, ggplot2_0.9.0 
library(stringr)
library(ggplot2)
library(plyr)
# reading previously cleaned data
load("data_files.rdata")
# Sourcing functions
source("functions.r")
# Setting up parameters
annual_mean <- 25
# mean value is in cm

Bad ✖
library(ggplot2)
foo <- read.csv("file1.csv")
a <- 1
b <- 3
test <- function(foo2) {
    return(a + b)
}

Quick guide to R data types

# Vector: (single dimension, all same type)
vec1 <- 1:10
class(vec1)
## [1] "integer"
vec2 <- letters[1:10]
class(vec2)
## [1] "character"
# Data Frame: Each column is a vector, but adjacent vectors can hold different things
# Matrix: Just like a data.frame except it's all numeric
# List: (any dimension, mix and match)
l1 <- list(A = data.frame(x = 1:10, y = rnorm(10)),
    B = 1, C = letters[1:3])
str(l1)
## List of 3
  $ A:'data.frame':	10 obs. of  2 variables:
   ..$ x: int [1:10] 1 2 3 4 5 6 7 8 9 10
   ..$ y: num [1:10] 0.618 0.519 0.343 0.428 -0.885 ...
  $ B: num 1
  $ C: chr [1:3] "a" "b" "c"

Finding Help

Locally

		?function_name
		??function_name
		RSiteSearch("function_name")
	

Online

StackOverflow.com/questions/tagged/r

Rseek.org

sessionInfo()

sessionInfo()
## R version 2.14.2 (2012-02-29)
 Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)

 locale:
 [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

 attached base packages:
 [1] graphics  grDevices utils     datasets  stats     methods   base

 other attached packages:
 [1] knitr_0.3      formatR_0.3-4  devtools_0.6   plyr_1.7.1
 [5] reshape2_1.2.1 ggplot2_0.9.0

 loaded via a namespace (and not attached):
  [1] codetools_0.2-8    colorspace_1.1-1   dichromat_1.2-4
  [4] digest_0.5.1       evaluate_0.4.1     gdata_2.8.2
  [7] grid_2.14.2        gtools_2.6.2       highlight_0.3.1
 [10] MASS_7.3-17        memoise_0.1        munsell_0.3
 [13] parser_0.0-14      proto_0.3-9.2      RColorBrewer_1.0-5
 [16] Rcpp_0.9.10        RCurl_1.91-1       scales_0.2.0
 [19] stringr_0.6        tools_2.14.2

Use dput() to share some data

dput(head(mtcars))
structure(list(mpg = c(21, 21, 22.8, 21.4, 18.7, 18.1), cyl = c(6,
 6, 4, 6, 8, 6), disp = c(160, 160, 108, 258, 360, 225), hp = c(110,
 110, 93, 110, 175, 105), drat = c(3.9, 3.9, 3.85, 3.08, 3.15,
 2.76), wt = c(2.62, 2.875, 2.32, 3.215, 3.44, 3.46), qsec = c(16.46,
 17.02, 18.61, 19.44, 17.02, 20.22), vs = c(0, 0, 1, 1, 0, 1),
     am = c(1, 1, 1, 0, 0, 0), gear = c(4, 4, 4, 3, 3, 3), carb = c(4,
     4, 1, 1, 2, 1)), .Names = c("mpg", "cyl", "disp", "hp", "drat",
 "wt", "qsec", "vs", "am", "gear", "carb"), row.names = c("Mazda RX4",
 "Mazda RX4 Wag", "Datsun 710", "Hornet 4 Drive", "Hornet Sportabout",
 "Valiant"), class = "data.frame")

Leverage your .rprofile

Set options

options(max.print = 2000)
options(prompt = "$ ")
options(stringsAsFactors = FALSE)
# Store API keys
options(MendeleyKey = "My_secret_key")

	See ?options for more information on settings
	options() to list current settings

Leverage your .rprofile

Load frequently used libraries

library(ggplot2)
library(stringr)
library(plyr)
library(devtools)

Leverage your .rprofile

Load custom functions

# A function that tells me which packages are out of date
check.packages <- function() {
    if (!is.null(utils::old.packages())) {
        old_packages <- utils::old.packages()
        cat("Notification:", dim(old_packages)[1], "packages are out of date \n")
        cat(unname(old_packages[, 1]), sep = ",", "\n")
    }
    if (is.null(utils::old.packages())) {
        cat("All packages are current \n")
    }
}

Word of caution regarding.rprofile

# To load R without the .rprofile
R -- vanilla

Thanks to Hadley for pointing this out oversight.

Reading Data

from flat text files

read.table

from databases

Use packages RODBC, RMySQL

from cloud storage

Amazon S3, Google Docs, Dropbox etc.

Saving Data

save(data, file = "slots.rdata")

write.table(data, file = "slots-3.csv", sep = ",",
    row = F)

Manipulating Data - Plyr

Hadley Wickham (2011). The Split-Apply-Combine Strategy for Data Analysis. JSS, 40(1), 1-29. for more details

ddply example

data <- data.frame(x = c("a", "a", "b", "b", "c",
    "c"), y = c(2, 4, 0, 5, 5, 10))

ddply(data, .(x), summarise, y = mean(y))

Plyr syntax

result <- ddply(data, variable, summarise, n = sum(n))

result <- llply(list_name, function_name)

reshape2 allows you to reshape data into any format possible

dcast(melted_data, temp ~ light, length)

dcast(melted_data, temp ~ light, mean)

dcast(melted_data, temp ~ light, custom_function)

Example of melting and casting

test_data <- data.frame(id = 1:9, category = factor(rep(sample(letters[1:3]),
    3)), treatment = rep(sample(c("control", "trt_1", "trt_2")),
    3), price_index = rnorm(9) * 200, prev_yr_index = rnorm(9) *
    200)

melted_data <- melt(test_data, id.vars = 1:3)

dcast(melted_data, category + treatment ~ variable,
    length)
dcast(melted_data, category + treatment ~ variable,
    mean)

Writing Functions

If you have to repeat the same 3-4 lines of code more than once, turn it into a function

Use a version control system like git

Dynamic report generation

	# As easy as:
	knit('report.rnw')
	

	# All the syntax in this talk was generated using knitr

R on the cloud

R scales really well on cloud platforms (AWS, most anything)

R can interact with any web service that has an API

Make function calls to R using REST API

opencpu.org

Questions?