\documentclass[xcolor = pdftex, dvipsnames,table] {beamer} \usetheme{Berkeley} \title[Introduction to R]{The Briefest of Introductions to R} \author{Lawrence Hubert} \institute{Department of Psychology \\The University of Illinois} \date{A Programming Language and Software Environment for Statistical Computing and Graphics} \begin{document} \begin{frame} \titlepage \end{frame} \begin{frame} \frametitle{The R Programming Language} Originally developed by Ross Ihaka and Robert Gentleman in the late 1990's at the University of Auckland, New Zealand. It is now developed by the \emph{R Development Core Team} \bigskip R is an implementation of the S programming language developed at Bell Laboratories by John Chambers and colleagues. Another implementation of S is in the commercially available product, S-Plus. \bigskip Extra Credit Question: What other languages and operating systems were developed at Bell Labs? \bigskip The name `R' comes from (partly) the first name of the two original authors, and as a letter-play on the name `S'. \end{frame} \begin{frame} \frametitle{} The main place to get everything you need about R is: \bigskip http://www.r-project.org \bigskip and one of the CRAN mirrors (Comprehensive R Archive Network): \bigskip http://cran.r-project.org \bigskip R is widely used for statistical software development and data analysis; it could now be considered the \emph{lingua franca} for statisticians, and is analogous to how MATLAB is viewed for the Engineering and Computer Science community. \end{frame} \begin{frame} \frametitle{} R is ``state of the art''; (almost all) statistical researchers provide their ``cutting edge'' methods as R packages. \bigskip The source code is freely available under the GNU General Public License; also, precompiled binary versions are available (and free) for Windows, Mac OS X, and Linux. \bigskip Generally, R uses a command line interface, but several GUIs are available. We will introduce one done by John Fox, called R Commander, towards the end of our session. \bigskip If you need a GUI to do any of the analyses discussed, say, in 406/7, then R Commander is for you. \end{frame} \begin{frame} \frametitle{} R supports a large variety of statistical and numerical techniques in its (eight) base packages (e.g., the base stats package); in fact, most standard methods (through all of standard multivariate analysis) are already available in this default installation. \bigskip R is also highly extensible through the use of packages --- user-submitted libraries for specific functions or specific areas of study. We will list some later that are (most) relevant to psychology. \bigskip Because of its S language lineage, R has stronger object-oriented programming facilities than most statistical computing languages. \end{frame} \begin{frame} \frametitle{} The term ``environment'' is intended to characterize R as a fully planned and coherent system, rather than as an incremental accretion of very specific and inflexible tools; this is frequently the case with other data analysis software (e.g., SAS, SPSS, SYSTAT, ...) \bigskip R (like S) is designed around a true computer language; it allows users to add additional functionality by defining new functions. Much of the system is itself written in R, making it easy for users to follow the algorithmic choices made. \bigskip For computationally-intensive tasks, C, C++, and Fortran code can be linked and called at run time. Also, advanced users can write C code to manipulate R objects directly. \end{frame} \begin{frame} \frametitle{} R is more than a statistics system. It is an environment within which statistical techniques are implemented. \bigskip R has its own Latex(-like) documentation format, which is used to supply comprehensive documentation, both on-line in a number of formats and in hardcopy. \bigskip Generally, a statistical package such as SPSS is oriented toward combining instructions and rectangular data sets to produce (voluminous) printout and graphs. Routine, standard data analysis is easy; innovative or nonstandard analysis is hard or impossible. \end{frame} \begin{frame} \frametitle{} A programming environment is oriented toward transforming one data structure into another. Programming environments, such a R (or S), are extensible; standard data analysis is easy, but so are innovation and nonstandard analysis. \bigskip One of R's strengths is its graphical capabilities, which produce publication-quality graphs that include mathematical symbols. \bigskip Although R is mostly used by statisticians and other practitioners requiring a complete environment for statistical computation and software development, it could also be used as a general matrix manipulation and calculation Toolbox. \end{frame} \begin{frame} \frametitle{} Such a calculator usage is much like MATLAB (and the free OCTAVE). This will not be our emphasis here, however. Generally, I prefer MATLAB for this one-off calculator-type task that deals explicitly with matrices. \bigskip Using R to do some matrix manipulations is fairly straightforward, and comparable to how MATLAB does things. You may need some of this facility, however, if you ever wish to write your own functions, and, ultimately, your own packages. \bigskip As documentation, see Bill Revelle's matrix\_algebra\_in\_r.pdf at the class web site. \end{frame} \begin{frame} \frametitle{Packages} The capabilities of R are extended by user-contributed \emph{packages} (comparable to Toolboxes in MATLAB), allowing specialized statistical techniques, graphical devices, as well as programming interfaces and import/export capabilities to many external data formats. \bigskip A core set of packages are included with the installation of R, with over 1000 more available at a CRAN site. \bigskip Notable packages are listed along with comments on the official R Task View pages: \bigskip http://cran.r-project.org/web/views \end{frame} \begin{frame} \frametitle{} Task views of particular interest are: Cluster; Environmetrics; Multivariate; Psychometrics; Social Sciences \bigskip The Bioinformatics community has started a successful effort to use R for the analysis of data from molecular biology laboratories. \bigskip The Bioconductor Project, started in 2001, provides R packages for the analysis of genomic data, such as Affymetrix and cDNA microarray object-oriented data handling and analysis tools. \bigskip http://www.bioconductor.org \end{frame} \begin{frame} \frametitle{} Jonathan Baron (U Penn, Psychology) has a very nice R help page. It gives a complete list of all packages, plus a search facility to look for what you might need: \bigskip http://finzi.psych.upenn.edu \bigskip Also, remember the R project page and the Cran mirrors: \bigskip http://www.r-project.org \bigskip http://www.cran.r-project.org \end{frame} \begin{frame} \frametitle{} The \emph{Journal of Statistical Software} is an on-line resource founded in 1996 by the American Statistical Association as a freely available and peer reviewed resource for statistical software and algorithms. \bigskip The site listed below for \emph{JSS}, is devoted to an open-source philosophy. Thus, for both articles and code snippets, the source code is published along with the paper: \bigskip http://www.jstatsoft.org \bigskip Although code implementations can use different languages or computing environments, the emphasis is mainly on R, and, to a lesser extent, MATLAB. \end{frame} \begin{frame} \frametitle{} Several other good sources of R related material: \bigskip \emph{R News} was (now, \emph{The R Journal}) the newsletter of the R project, and features short to medium length articles covering topics that might be of interest to users or developers of R. It is all free, and on-line in pdfs: \bigskip http://journal.r-project.org \bigskip Bill Revelle (Northwestern, Psychology) maintains the Personality Project, and a very good introduction to R for psychological research: \bigskip http://www.personality-project.org/r/r.guide.html \end{frame} \begin{frame} \frametitle{Some Contributed Packages of Particular Interest in Psychology} See Rnews (December, 2007) on the Psychometrics Task View; and meta: An R Package for Meta-Analysis. \bigskip See the R package Zelig (Everyone's Statistical Software) --- \bigskip Extra Credit Question: Who was Zelig and what actor played him? \end{frame} \begin{frame} \frametitle{} Multilevel Modeling: multilevel and nlme: \bigskip http://cran.r-project.org/doc/contrib/Bliese\_Multilevel.pdf \bigskip Neural Networks: nnet and AMORE (A MORE flexible neural network package) \bigskip coin: A computational framework for conditional inference (i.e., think exact tests; permutation tests; randomization tests) \bigskip tsp: Infrastructure for the Traveling Salesperson Problem \bigskip tree: Classification and regression trees \end{frame} \begin{frame}{} seriation : Row/column object ordering in proximity matrices \bigskip sem: Structural equation models \bigskip rimage: Image processing module for R \bigskip mlica: Maximum likelihood implementation of Independent Components Analysis \bigskip lsa: Latent semantic analysis (using document-term matrices) \bigskip labdsv: Ordination (i.e., scaling) and multivariate analysis for ecology \end{frame} \begin{frame} \frametitle{} gap: Genetic analysis package \bigskip fmri: Analysis of fMRI experiments \bigskip AnalyzeFMRI: Functions for analysis of fMRI data sets stored in the ANALYZE or NIFTI format \bigskip ecodist: Dissimilarity-based functions for ecological analysis (contains nmds for non-metric multidimensional scaling) \bigskip eba: Elimination-by-aspects models \bigskip lp: Interface to Lp\_solve for linear/integer programs \end{frame} \begin{frame} \frametitle{} Rglpk: Linear and mixed integer programming solver using GLPK (GNU Linear Programming Kit) \bigskip clValid: An R package for cluster validation (also, see the CRAN Task View on Cluster Analysis and Finite Mixture Models) \bigskip clue: cluster ensembles (contains a lot of my work on fitting ultrametrics and additive trees by least-squares iterative projection) \bigskip qgen: Quantitative genetics using R \bigskip psyphy: Functions for analyzing psychophysical functions \end{frame} \begin{frame} \frametitle{} smacof: multidimensional scaling (from Jan de Leeuw) \bigskip psych: Procedures for personality and psychology research (from Bill Revelle) \bigskip ade4: Analysis of ecological data: Exploratory and Euclidean methods in environmental sciences \bigskip anacor: Simple and canonical correspondence analysis \bigskip amap: Another multidimensional analysis package \bigskip acepack: ace and avas for selecting regression transformations \end{frame} \begin{frame} \frametitle{} e1071: Packages from Vienna, including svm (support vector machines) and lca (latent class analysis) \bigskip cba: Clustering for business analytics (includes order.optimal for ordering the leaves of a tree --- one of my research areas as well) \bigskip MASS: Package for the text \emph{Modern Applied Statistics with S} (contains isoMDS to carry out Kruskal's non-metric multidimensional scaling) \end{frame} \begin{frame} \frametitle{Packages for Social Network Analysis} sna: Tools for social network analysis \bigskip network: Classes for relational data \bigskip latentnetHRT and latentnet : Latent position and cluster models for networks \bigskip inetwork: Network analysis and plotting \bigskip statnet: A suite of packages for network analysis: \bigskip http://csde.washington.edu/statnet \bigskip ergm: Exponential-family models for networks \end{frame} \begin{frame} \frametitle{} degreenet: Models for skewed count distributions relevant to networks \bigskip Also, see STOCNET at \bigskip http://stat.gamma.rug.nl/snijders/ \bigskip This is an open software system for the advanced statistical analysis of social networks (primarily, the work of Tom Snijders). \end{frame} \begin{frame} \frametitle{Gnumeric} The Gnumeric spreadsheet is part of the GNOME (Linux) desktop environment. GNOME is a project to create a free, user friendly desktop environment for Linux. \bigskip The goal of Gnumeric is to be the best possible spreadsheet. It will import existing Microsoft Excel files, and it won't screw-up simple statistical computations (as Excel routinely does; Microsoft seems incapable of coming up with a quality product --- this is continually documented in the statistical literature). \end{frame} \begin{frame} \frametitle{} There is an open-source build for Windows, as well as the various Linux versions. See: \bigskip http://www.gnome.org/projects/gnumeric \bigskip Gnumeric can even do linear and mixed integer programming. \bigskip It is an explicit friend of R. \end{frame} \begin{frame} \frametitle{Using R} One strong characteristic of R is the great help system. Try: \bigskip help(t.test) \bigskip help.search("cluster") \bigskip Also, the various manuals (involving a huge number of pages) are available in pdfs when you install R. The various contributed packages also come with various kinds of help documentation. \end{frame} \begin{frame} \frametitle{} Before we can use R or any of the contributed programs, we need to get our data (usually, a subject by variable rectangular matrix) into what is called a \emph{data frame}. \bigskip Here are a number of ways of doing this: \bigskip 1) Keyboard input \bigskip 2) Reading data into a data frame from a textfile \bigskip 3) Using the spreadsheet-like data editor in R \bigskip 4) Importing data from some spreadsheet format (e.g., *.xls), or from SPSS (e.g., *.sav) \bigskip 5) Accessing data that are already in R libraries \end{frame} \begin{frame} \frametitle{} We have our community data in a textfile called communitydata.txt , with verbatim contents as follows: \smallskip accidents vehicles police a 1 4 20 b 4 10 6 c 5 15 2 d 4 12 8 e 3 8 9 f 4 16 8 g 2 5 12 h 1 7 15 i 4 9 10 j 2 10 10 \end{frame} \begin{frame} \frametitle{Some Commands to Try} accidents = c(1,4,5,4,3,4,2,1,4,2) \bigskip comm.labels = c('a','b','c','d','e','f','g', 'h','i','j') \bigskip vehicles = scan() \bigskip 4 10 15 12 8 16 5 7 9 10 \bigskip police = scan() \bigskip 20 6 2 8 9 8 12 15 10 10 \bigskip community = data.frame(comm.labels,accidents,vehicles,police) \end{frame} \begin{frame} \frametitle{} community \bigskip community.altone = edit(as.data.frame(NULL)) \bigskip community.altone \bigskip fix(community) \end{frame} \begin{frame} \frametitle{} community.alttwo = read.table('G:/r\_class\_material/communitydata.txt', header = TRUE) \bigskip community.alttwo \end{frame} \begin{frame} \frametitle{} attach(community) \bigskip summary(community) \bigskip community.model = lm(accidents $\sim$ vehicles + police) \bigskip summary(community.model) \bigskip plot(community.model) \bigskip \end{frame} \begin{frame} \frametitle{Accessing and Using Data in Libraries} Assuming that a package has been ``installed'': \bigskip install.packages('car') \bigskip library(car) \bigskip data(Prestige) \bigskip attach(Prestige) \bigskip objects() \bigskip search() \end{frame} \begin{frame} \frametitle{} dim(Prestige) \bigskip mean(income) \bigskip summary(type) \bigskip summary(education) \end{frame} \begin{frame} \frametitle{} plot(income,prestige) \bigskip abline(lm(prestige $\sim$ income)) \bigskip title('Scatterplot of Prestige vs. Income') \bigskip identify(income,prestige,row.names(Prestige)) \end{frame} \begin{frame} \frametitle{User Defined Functions} Functions are the core of R. They take inputs and produce outputs. \bigskip hello = function(x) \textbraceleft \bigskip cat("hello",x,"$\backslash$n") \bigskip \textbraceright \bigskip hello("and goodbye") \end{frame} \begin{frame} \frametitle{} absolute = function(x) \textbraceleft \bigskip if (x $<$ 0) \textbraceleft \bigskip return (-x) \textbraceright \bigskip else \textbraceleft \bigskip return(x) \bigskip \textbraceright \bigskip \textbraceright \bigskip absolute(-5) \end{frame} \begin{frame} \frametitle{} factorial = function(x) \textbraceleft \bigskip ret = 1 \bigskip for(i in 1:x) \textbraceleft \bigskip ret = ret*i \bigskip \textbraceright \bigskip return(ret) \bigskip \textbraceright \bigskip factorial(5) \end{frame} \begin{frame} \frametitle{} factorial.alt = function(x) \textbraceleft \bigskip i = 1 \bigskip f = 1 \bigskip while (i $< =$ x) \textbraceleft \bigskip f = f*i \bigskip i = i + 1 \bigskip \textbraceright \bigskip f \textbraceright \bigskip factorial.alt(5) \end{frame} \begin{frame} \frametitle{} ourhistogram = function(x,breaks = "Scott", col = "purple",...) \textbraceleft \bigskip hist(x,breaks = breaks, probability = TRUE, col = col, ...) \bigskip \textbraceright \bigskip x = rnorm(200) \bigskip ourhistogram(x,xlab = "histogram of x", col = "green") \bigskip \end{frame} \begin{frame} \frametitle{R Commander} R Commander is a Basic Statistics GUI for R, written by John Fox from McMasters (Sociology) \bigskip install.packages("Rcmdr",dependencies = True) \bigskip library(Rcmdr) \end{frame} \begin{frame} \frametitle{RWinEdt} RWinEdt is a package that provides a plug-in for using WinEdt as an editor for R. This gives syntax highlighting, indenting, and so forth. \bigskip Remember the shareware license information on WinEdit that we have in our Latex presentation. The Psychology Department has a site license. \bigskip Also, you might check out Tinn-R. \end{frame} \end{document}