Intro to R. Some history. Some history

Size: px

Start display at page:

Download "Intro to R. Some history. Some history"

Scott Berry
5 years ago
Views:

Intro to R Héctor Corrada Bravo CMSC858B Spring 2012 University of Maryland Computer Science http://www.nytimes.

com/forbes/2010/0524/opinions-software-norman-nie-spss-ideas-opinions.html

uk/2010/05/06/revolution_commercial_r/ Some history John Chambers and others started developing the S language in 1976 Version

1 Intro to R Héctor Corrada Bravo CMSC858B Spring 2012 University of Maryland Computer Science Some history John Chambers and others started developing the S language in 1976 Version 4 of the language definition (currently in use) was settled in 1998 That year, S won the ACM Software System Award Commercially available as S-PLUS Some history Ihaka and Gentleman (of NYTimes fame) create R in 1991 They wanted lexical scoping (see NYTimes pic) Released under GNU GPL in 1995 Maintained by R Core Group since1997

Currently Freely available: http://www.r-project.org/ IDEs: [cross-platform] http://rstudio.org/ [Windows and Linux] http:// www.revolutionanalytics.

2 Currently Freely available: IDEs: [cross-platform] [Windows and Linux] Languages used in Kaggle (prediction competition site) Also bindings for emacs [ and plugin for eclipse [ Resources: Manuals from r-project manuals.html Chambers (2008) Software for Data Analysis, Springer. Venables & Ripley (2002) Modern Applied Statistics with S, Springer. List of books: Programming language features: Conceptually, it is very similar to Scheme (functional, lexical scope, lists are basic data structure) with saner syntax. Syntax is nice for numerical computation (similar to matlab) Many language features there to support statistical modeling (formula syntax, data.frames) Objects (we ll see that next time with Bioconductor) Fairly clean C interface (non-base package Rcpp provides awesome interface to C++) Interactive (REPL), but also scripting available

3 Uses a package framework (similar to Python) Documentation system: > help( sapply )# bring up help page >?sapply # shortcut >??sapply # search for string in docs > help.start()# open doc index Divided into two parts base: what you get when you download R (base package, and other packages like stats, graphics, utils, Matrix, boot,codetools) everything else: Follower Map loading packages > library(twitter) Asia 4 Africa 1 N. America 79 S. America 7 Australia/N.Z. 1 Europe 17 > require(twitter) installing packages > install.packages( twitter ) Created twittermap

4 Function definition twittermap <- function(username, userlocation=null, #default value filename="twittermap.pdf", nmax = 1000) { } <body> Function call twittermap("yankees", userlocation="new York", nmax=500,file="yanks.pdf") Lists are basic data structure (like scheme) # creating a list (with names) > list(users=followers, location=followerslocation) # accessing named element > tmp$users # accessing element by index > followers[[1]] # slicing list > followers[1:3] Functional language followerslocation = sapply(followers,location) Equivalent (really bad idea in general) followerslocation = c() for (i in 1:length(followers)) { followerslocation[i]=location(followers[[i]]) } vectors # creating vec = c(1,10,20) vec = 1:100 vec = seq(1,100,by=2) vec = rnorm(100) # indexing vec[1] vec[1:10] # operations sum(vec) mean(vec) vec/10 crossprod(vec) tcrossprod(vec)

5 Matrices # creating mat = matrix(c(1,10,20,30), nrow=2, ncol=2) mat = matrix(rnorm(100), nrow=20, ncol=5) # indexing mat[1,1] # element in row 1 column 1 mat[,1] # column 1 # operations sum(mat) # sum of all entries colsums(mat) # column-wise sum apply(mat,2,sum) # same thing rowmeans(vec)# row-wise means # operations with vectors and scalars mat/10 # divide all entries by scalar vec = runif(20) mat/vec # divide each column by vec vec = rnorm(5) sweep(mat,2,vec, / ) # divide each row by vec Data frames: a hybrid of matrix and list # creating (looks like a named list) distdf=data.frame(dist=c(yanksll[,3], natsll[,3]), team=rep(c( yanks, nats ), c(nrow(yanksll), nrow(natsll))) # accessing # like a list distdf[[1]] # the first element (column) distdf$dist # a named element (column) names(distdf) # the names of elements (columns) # like a matrix distdf[1,1] # the first value in first column distdf[,1] # the entire first column Formula syntax: statistical modeling is part of the language # a linear regression model # fit = lm (dist~team, data=distdf) formula # which you can get information about summary(fit) Plotting: there are three graphics system in R: graphics: the base system (which we ll use today) lattice: a very flexible system (uses statistical model syntax we ll see later) ggplot2: very pretty, very extensible (grammar of graphics) R graph gallery:

6 Statistical tests available out of the box as objects you can compute on # perform a KS test for differences in # distribution of distances testres=ks.test( log(subset(distdf,team=="yankees")$dist+1), log(subset(distdf,team=="nationals")$dist+1), alternative="greater") # print result testres # get value of test statistic testres$stat # get P-value for test testres$p.value

An Introduction to R- Programming

An Introduction to R- Programming Hadeel Alkofide, Msc, PhD NOT a biostatistician or R expert just simply an R user Some slides were adapted from lectures by Angie Mae Rodday MSc, PhD at Tufts University