SIMPLE TUTORIAL in R

R download from http://www.r-project.org

partially based in http://www.ats.ucla.edu/stat/r/notes

# INTRODUCTION # R is accused of being slow, memory-hungry, and able # to handle only small data sets. This is completely true. # Fortunately, computers are fast and have lots of memory. Data # sets with a few tens of thousands of observations can be handled # in 256Mb of memory, and quite large data sets with 2Gb of memory. #......................................................................... # BASICS # Lines with comments in R begin with sign # help(solve) ?solve # information on any specific named function # alternative # for help pages related to "plot" # for functions whose names contain "plot"

help.search("plot") apropos("thing")

# We can search any function in http://www.r-project.org #......................................................................... # OBJECTS and BASICS # We can store data in an "object": N <- 1000 N = 1000 # alternative y <- c(3.1,10.5,14,30,15,19) x <- c(4,12,12,20,16,22) strata <- c("Madrid","Barcelona","Lisboa") ok.set <- c(T,T,F,T,F) ls() # character values # logical values

# display the names of the objects in the workspace

# If you type the "object name" you see what is stored in the object N y x # To see what objects have been created objects()

# To remove an object rm(x) rm(y,z) # Other ways to enter/create data z <- seq(1,10) z <- c(rep(3,4),rep(5,2)) # a sequence of values # join sequences of values

majors <- c(rep("Forestry",3),rep("Fisheries",5),rep("Math",2), "Education",rep("Business",2))

setwd("C:/kk") getwd()

# set to wherever your data directory is located # check that you are in the correct directory

# Run an ascii program written in R source("c:/.../program.R") #......................................................................... # READ in DATA from a DATA FILE # The easiest format in a file has variable names in the first row: # case id gender deg # 1 1 F Other # 2 2 M Other # 3 2 M Other # 4 4 M PhD yrdeg 92 91 91 96 field startyr year rank admin Other 95 95 Assist 0 Other 94 94 Assist 0 Other 94 95 Assist 0 Other 95 95 Assist 0

# and fields separated by spaces. salary <- read.table("c:/.../salary.txt", header=TRUE) # Data from the file salary.txt are stored into the data frame # object "salary". # # # # # # # # HINTS: Many statistical packages (SAS, SPSS) can save data as an EXCEL file. Import any type of data into R by using EXCEL and saving there the data file into a comma delimited (*.csv) format. Once the comma delimited file is created using the "Save As" feature in EXCEL you can import it into R using either the read.table() or the read.csv() function.

thing <- read.table("c:/.../myfile.csv", "header=T", sep=",") # Alternatively, you can use read.csv() thing <-read.csv("c:/.../myfile.csv","header=T") # Alternatively, tou can use the clipboard: # Open the *.xls file in EXCEL # Select and copy the relevant cells in Windows thing <- read.table(file="clipboard",sep="\t",header=T) # The file "clipboard" instructs read.table to read the file from the Windows # clipboard, and the separator option of "\t" notifies read.table that elements

# are separated by tabs. # The same way form R to EXCEL: # Going to EXCEL and issuing the "paste" command will put the matrix # into the EXCEL worksheet. write.table(mymatrix,file="clipboard",sep="\t",col.names=NA) # Files for read.table can also 'live' on the web fl2000 <- read.table("http://faculty.washington.edu/tlumley/ data/FLvote.dat", header=TRUE) # Another type of commonly used ASCII data format is fixed format. # In this format data are placed in a fixed column for each observation. # It requires a codebook to specify which column corresponds to which variable. # # # # # # # # # # # # # # # # # # # # # # # # # # Example: data are in file 'datfix.txt: 195 094951 26386161941 38780081841 479700 870 56878163690 66487182960 786 069 0 88194193921 98979090781 107868180801 variable name column number id 1-2 a1 3-4 t1 5-6 gender 7 a2 8-9 t2 10-11 tgender 12 To read these data we use the read.fwf() function on fixed format data instead of the read.table() function. Here, we use the width argument which indicates the width of each variable instead of using the sep argument to indicate the start of each variable.

fixed <- read.fwf("schdat_fix.txt", width = c(2, 2, 2, 1, 2, 2, 1)) names(fixed) <- c("id", "a1", "t1", "gender", "a2", "t2", "tgender") fixed # check the data

# Sometimes we read data from other packages, such as Stata or SPSS. library(foreign) # library to read foreign datasets # read.dta: read Stata (.dta) data files # read.spss: read SPSS (.sav) data files #...............................................................................

# SIMPLE ARITHMETIC OPERATIONS x+1 x+y 5*x x^2 sqrt(x) log(x) # # # # # # add a 1 to all values in x add x and y multiply all values in x by 5 take the square of all values in x take the square root of each value in x take the natural log of each value in x

# Example: a sequence of arithmetic operations instead of one step xbar <- mean(x) diffs <- (x - xbar) # subtract mean of x from each value diffs.sq <- diffs^2 # square all the differences ssx <- sum(diffs.sq) # this is Sum of Squares of X ssx <- sum((x-mean(x))^2) # can be done in one step #............................................................................... # LISTS and DATA.FRAMES # Examples of lists w <- list(strat1=c(3,2,3),strat2=c(8,10,12,15)) x <- list(people=c("Zoe","Rapunzel","Igor"), state=c("AK","AL","AK"),age=c(20,28,98)) # # # # Example: One way to make side-by-side boxplots: make a list of the values with each component in the list corresponding to a different sample

y <- list(sample1=c(18,12,9,7,15,20),sample2=c(18,11,12,22,23,30), sample3=c(35,42,32,37,41,41,38,39)) boxplot(y) # OPERATIONS with lists x <- list(one=c(18:36),two=c("AK","AL","AZ"), three=c(T,T,F,T),four=matrix(1:12,3,4)) # Access to components x[[1]] # by order x$one # by name # Access to elements within components x[[1]][3:6] x$one[3:6] unlist() unlist(x) # convert a list to a vector # handy for printing out returned values from function

# List version of apply is lapply() # (see next item of matrices) lapply(x,length) # DATA FRAMES: a special kind of list object; number of elements must # be the same for all components muscle <- rnorm(n=10,mean=3,sd=1) sex <- factor(rep(c("M","F"),c(6,4))) speed <- rep(0,10) speed[1:6] <- rnorm(6,30-2*muscle[1:6],2)

speed[7:10] <- rnorm(4,40-2*muscle[7:10],2) mydata <- data.frame(y=speed,x1=muscle,x2=sex) mydata # Dealing with variables # # # # # # # # Commands: rbind: combines rows of data merge: match merges two data frames dimnames: lists or assigns names of data frames cbind: combines columns of data sapply: applies a function to elements of a list factor: creates a categorical variable with value labels if desired table: creates frequency table

# Keeping and Dropping Variables hs1 <- read.table("http://www.ats.ucla.edu/stat/R/notes/hs1.csv", header=T, sep=",") attach(hs1) # Keeping only the observations where the reading score is 60 or higher. hs1.read.well <- hs1[read >= 60, ] # Comparing means of read in the original hs1 data frame and the # new smaller hs1.read.well data frame. mean(hs1.read.well$read) mean(hs1$read) # Keeping only the variables id, female, read and write from the # hs1.read.well data frame. names(hs1.read.well) hs1.kept <- hs1.read.well[ , c(1, 2, 7, 8)] names(hs1.kept) # Dropping the variables ses and prog from the hs1.read.well data frame names(hs1.read.well) hs1.drop <- hs1.read.well[ , -c(4, 12)] names(hs1.drop) detach() # # # # Consider two files: hsmale.txt with the information for the males hsfemale.txt with the information for the females Combine these two files

hsfemale <- read.table('http://www.ats.ucla.edu/stat/R/notes/hsfemale.txt', header=T, sep=",") hsmale <- read.table('http://www.ats.ucla.edu/stat/R/notes/hsmale.txt', header=T, sep=",") table(hsfemale$female) table(hsmale$female) # Use the rbind function when we stack data because we combine rows of data hsmasters <- rbind(hsfemale, hsmale) table(hsmasters$female) detach() # Merge two data frames on a variable (or a list of variables). # We use variable id which has the same name in both data sets. # Specifying T in the all argument indicates that we want to keep

# all the observations from each data set rather than only keeping # the observations that came from both data sets. hsdem <- read.table('http://www.ats.ucla.edu/stat/R/notes/hsdem.txt', header=T, sep=",") hstest <- read.table('http://www.ats.ucla.edu/stat/R/notes/hstest.txt', header=T, sep=",") hsdem hstest hsdiss <- merge(hstest, hsdem, by="id", all=T) hsdiss # If the variable that we were merging on had different names in each # data frame then we could use the by.x and by.y arguments. # # # # In the by.x argument we would list the name of the variable(s) that was in the data frame listed first in the merge function (in this case in hstest) and in the by.y argument we would name the variable(s) that was in the data frame listed second (in this case hsdem).

hsdiss.1 <- merge(hstest, hsdem, by.x="id", by.y="id", all=T) hsdiss.1 # Other option by creating an indicator of which data set the observations # came from from <- data.frame(rep(1, length(hsdem$id))) dimnames(from)[[2]] <- "from" hsdem.1 <- cbind(hsdem, from) hsdem.1 from <- data.frame(rep(1, length(hstest$id))) dimnames(from)[[2]] <- "from" hstest.1 <- cbind(hstest, from) hstest.1 hsdiss.2 <- merge(hstest.1, hsdem.1, by.x="id", by.y="id", all=T, suffix=c("test", "dem")) attach(hsdiss.2) hsdiss.2$both[!is.na(fromtest) & !is.na(fromdem)] <- "both" hsdiss.2$both[is.na(fromtest)] <- "dem" hsdiss.2$both[is.na(fromdem)] <- "test" hsdiss.2 # Factor variables hs0 <- read.table("http://www.ats.ucla.edu/stat/R/notes/hs0.csv", header=T, sep=",") attach(hs0) # Check if any of the variables in the hs0 data frame are factor variables. sapply(hs0, is.factor) # Creating a factor (categorical) variable called schtyp.f for # schtyp with value labels. schtyp.f <- factor(schtyp, levels=c(1, 2), labels=c("public", "private")) search() detach() attach(hs0)

# Checking the factor variable schtyp.f in a frequency table. table(schtyp.f) schtyp.f # Creating a factor variable called female from gender with value labels. female <- factor(gender, levels=c(0, 1), labels=c("male", "female")) detach() attach(hs0) # Checking the factor variable female in a frequency table. table(female) table(race) race[race==5] <- NA detach() attach(hs0) table(race) # Creating a variable called total = read + write + socst total <- read+write+socst detach() attach(hs0) mean(total) # Creating a variable called grade based on total grade <- 0 grade[total >= 80 & total < 110] <- 1 grade[total >= 110 & total < 140] <- 2 grade[total >= 140 & total < 170] <- 3 grade[total >= 170] <- 4 detach() attach(hs0) table(grade) # Creating a factor variable called grade.f based on grade grade.f <- factor(grade, levels=0:4, labels=c("F", "D", "C", "B", "A")) detach() attach(hs0) is.factor(grade.f) table(grade.f) # Labels are nice when looking at frequency tables. table(schtyp, gender) # without labels table(schtyp.f, female) # with labels detach() #............................................................................... # OPERATIONS with VECTORS

# Examples of vectors with different types of "elements" w x y z <<<<c(3,2,1) c(T,T,F,F) c("Jane","Jill","Jeff","Matt") matrix(c(3,3,2,4,2,1),nrow=3,ncol=2) # # # # numeric valued logical valued character valued numeric valued matrix

# Accessing elements of a vector in 1 of 4 ways y <- c(18,32,15,-7,12,19) # Position in vector as positive integer

y[3:5] # Excluding elements, position as negative integers y[-c(1,5,6)] # By element name names(y) <- c("Joe","Bill","Karen","Helen","Ray","Paul") y[c("Helen","Ray")] # By logical conditions y[y<15] # Merging vectors # cbind() combines vectors by columns c1 <- c(10,20,30,40) c2 <- c(5,10,15,20) x <- cbind(c1,c2) x # rbind() combines vectors by rows x <- rbind(c1,c2) x #............................................................................... # OPERATIONS with MATRICES

y <- c(18,32,15,-7,12,19) x <- matrix(data=y,nrow=2,ncol=3)

# fill by columns first: it is the default # fill rows first

x <- matrix(data=y,nrow=2,ncol=3,byrow=T)

dimnames(x) <- list(c("r1","r2"),c("a","b","c")) apply(x,1,sum) apply(x,2,sum) apply(x,1,min) # sum across the 1st dimension, namely rows # sum across the 2nd dimension, columns

# Examples: A <- matrix(c(1, -2, 3, 4, -5, -6, 7, 8, 9, 0, 0, 10), 4, 3, byrow=TRUE) A t(A) diag(A) sum(diag(A)) # transpose a matrix # diagonal matrix # trace of a matrix

B <- matrix(c(-5, 1, 3, 2, 2, 6, 7, 3, -4), 3, 3, byrow=TRUE) A+B A-B -A # Product of matrices A %*% B

B %*% A # Inverse of a matrix: solve() # Example: A <- matrix(c(2, 5, 1, 3), 2, 2, byrow=TRUE) solve(A) # Check the result: A %*% solve(A) solve(A) %*% A det(A) eigen(A) # determinant of a matrix # eigenvalues and eigenvectors

#............................................................................... # CONDITIONS AND LOOPS # # # # # # if (...condition....) { ...code 1... } else { ...code 2... }

# while (...condition....) # {...code...} # for(rank of indices) # {...code...}

# Example 1 x <- 10 y <- 2 if (y >1){ x <- 2*x y <- 2*y } else{ x <- 38 x <-2*x } x y # Example 2 cunt <- c(0,0,0,0) n <- c(2,4,6,4) for(i in 1:length(n)){ cunt <- c(cunt,rep(i,n[i])) } cunt # Example 3 for (i in 1:10) print(i) n <- 10 while (n > 0) { cat(n,"is greater than 0 \n") n <- n - 1 }

#............................................................................... # USEFUL FUNCTIONS x <- c(10.1, 9.9, 11.2, 4.15, 2.3) prod(x) cumsum(x) diff(x) round(x,1) sort(x) rev(1:12) rank(x) # # # # # # # Product of vector elements Cumulative sums products Lagged differences Rounding of numbers Sorting or ordering vectors Reverse elements Sample ranks

# Example: find a minimum of a function x <- seq(0,5,0.001) fx <- x^3-8*x-20 m <- order(fx) fx[m[1]] x[m[1]] # # # # Samples To take a sample of a specified size from the elements of x using either with or without replacement sample(x, size, replace, prob)

# Example x <- 1:12 sample(x) sample(x,replace=TRUE) # a random permutation # bootstrap sampling (for length(x) > 1)

#............................................................................... # EXTEND THE LANGUAGE BY WRITING YOUR OWN FUNCTIONS # namefunction <- function(args) # { # ... code ... # } # y x z Examples of functions: <- c(3.1,10.5,14,30,15,19) <- c(4,12,12,20,16,22) <- cbind(x,y)

sd <- function(x) sqrt(var(x)) sd(x) circle.area <- function(radius) { area <- pi*radius^2 return(area) } circle.area(4) mystudy <- function(x){ par(mfrow=c(3,1)) hist(x[,1]) hist(x[,2]) plot(x[,1],x[,2]) par(mfrow=c(1,1))

10

apply(x,2,summary) } mystudy(z) #............................................................................... # SIMPLE STATISTICS, SUMMARIES, and PLOTS # # # # # # # # # # # # # # # # # # # # # # # # # Typical R functions: head: sapply: colMeans: colSums: rowSums: median: length: var: sd: tapply: cbind: summary: hist: histogram: boxplot: bwplot: stem: barplot: table: cor: lm: plot: abline: display first n observations applies a function to elements in a list column means column sums row sums calculates the median calculates the count calculates the variance calculates the standard deviation applies a function to each cell of a ragged array combining columns generic function provides a synopsis of an object histogram plot trellis histogram plot(s) box plot trellis box plot(s) stem-and-leaf plot bar plot frequency table calculates correlations fits a linear model generic plot function adds a line to an existing plot

# Example hs0 <- read.table("http://www.ats.ucla.edu/stat/R/notes/hs0.csv", header=T, sep=",") attach(hs0) hs0[1:20, ] names(hs0) vars <- hs0[ , 7:10] head(vars, n=10)

# shorthand way of referring to read, write, math, science

# The na.rm=T argument for the mean function is used to remove missing # observations from the computation of the means. sapply(hs0, mean, na.rm=T) sapply(vars, length) # count # the count for science is wrong, we create a new variable with only # the nonmissing cases of science and then use the length function science.good <- na.omit(science) length(science.good) sapply(vars, sapply(vars, sapply(vars, sapply(vars, median, na.rm=T) var, na.rm=T) sd, na.rm=T) min, na.rm=T) # median # variance # standard deviation

11

sapply(vars, max, na.rm=T) # Tukey's five number summary # - the maximum value # - the 75th percentile # - the 50th percentile # - the 25th percentile # - the minimum value sapply(vars, fivenum, na.rm=T) # We can also use the colMeans function to obtain the mean. # We can specify the variables by their numbers as in the sapply # or as variable names using cbind. colMeans(vars, na.rm=T) # Descriptive statistics can also be computed for a subset of the data frame: # we are looking at # who had a reading sapply(vars[read >= sapply(vars[read >= the summary statistics for only those students score of 60 or higher. 60, ], mean, na.rm=T) 60, ], median, na.rm=T)

# Obtaining the means of the variables write and science broken down by prgtype. # Science is the only variable with missing observations and thus # we use the na.rm to remove the missing observation. tapply(write, prgtype, mean) tapply(science, prgtype, mean, na.rm=T) tapply(write, tapply(write, tapply(write, tapply(write, prgtype, prgtype, prgtype, prgtype, length) var) sd) median) # # # # count variance standard deviation median

# Descriptive statistics for write by prgtyp in a much nicer display. m <- tapply(write, v <- tapply(write, med <- tapply(write, n <- tapply(write, sd <- tapply(write, cbind(mean=m, var=v, prgtype, mean) prgtype, var) prgtype, median) prgtype, length) prgtype, sd) std.dev=sd, median=med, n=n)

# More descriptive statistics including quantiles can be obtained by # using the summary function. summary(science) #............................................................................... # EXPLORING THE DATA THROUGH GRAPHS library(lattice) hist(write) # load trellis graphics

# trellis graphs histogram(~write, hs0, type="count") histogram(~write | gender, hs0, type="count")

# histogram of write by gender

# Note: In R it is possible to change the number of bins by # using the breaks argument in the hist function. hist(write, breaks=15)

12

# Put several plots on one image par(mfrow=c(2,1)) hist(write, breaks=15) hist(write) # boxplot of the variable write boxplot(write) # trellis graph of write by ses bwplot(ses ~ write, hs0) # trellis graph of boxplots of write by ses for each level of gender bwplot(ses ~ write| gender, hs0) # The graph shows ses by gender where the levels of ses are stacked # on top of another barplot(table(ses,gender), legend=c("low","medium","high"), ylim=c(0,135)) barplot(table(ses,gender), beside=T, legend=c("low","medium","high"), ylim=c(0,60)) #............................................................................... # FREQUENCY TABLES table(ses) # The table of write shows that it is generally undesirable to # obtain frequencies of continuous variables. table(write) table.vars <- hs0[ , c(1,5,6)] and prgtype sapply(table.vars, table) # shorthand way of referring to gender, schtyp

# Crosstabulation of gender and ses. tab1 <- table(gender,ses) tab1 # Compute the row and column proportions and frequencies # and a chisquare test of independence for the two-way table. prop.table(tab1,1) prop.table(tab1,2) rowSums(tab1) colSums(tab1) summary(tab1) # # # # # row proportions column proportions row frequencies column frequencies chi-square test of independence

# Correlations of write, read, math and science with listwise deletion # of missing values. # The correlations will not be calculated if there are missing values cor(vars, use="complete.obs") #...............................................................................

13

# ANALYZING DATA hs1 <- read.table("http://www.ats.ucla.edu/stat/R/notes/hs1.csv", header=T, sep=",") attach(hs1) # # # # # # # # # # # # # t.test: t-tests, including one sample, two sample and paired tapply: applies a function to each cell of a ragged array var: calculates the variance lm: fits a linear model (regression) anova: extracts the anova table from a lm object summary: generic function provides a synopsis of an object fitted: extracts the fitted values from a lm object resid: extracts the residuals from a lm object abline: generic function which adds a line to an existing plot glm: logistic regression drop1: compares model by dropping terms one at a time wilcox.test: non-parametric analyses kruskal.test: non-parametric analyses

# t-tests # one-sample t-test, testing whether the sample of writing scores # was drawn from a population with a mean of 50. t.test(write, mu=50) # paired t-test, testing whether or not the mean of write # equals the mean of read. t.test(write, read, paired=TRUE) # two-sample independent t-test. # use the tapply function to look at the variances of the variable # write for each group of female. tapply(write, female, var) t.test(write~female, var.equal=TRUE) # assuming equal variances t.test(write~female, var.equal=FALSE) # assuming unequal variances # ANOVA # # # # # # In R you can use either the aov function or the anova function combined with the lm function. The anova function extracts the anova table from the linear model fitted by the lm function. The aov function only fits an anova model and we use the summary function to see all the output.

anova(lm(write~factor(prog))) # is equivalent to summary(aov(write~factor(prog))) # two factors with interactions anova(lm(write~factor(prog)*female)) summary(aov(write~factor(prog)*female)) # Analysis of covariance (ANCOVA) # here, prog is the categorical predictor and read is the continuous covariate anova(lm(write~factor(prog) + read)) summary(aov(write~factor(prog) + read))

14

# REGRESSION summary(lm(write~female+read)) # # # # plot function will produce multiple diagnostic plots when applied to an lm object. These plots include residual versus fitted plots, qqplots of the residuals as well as scatter plots with the regression line overlaid

lm2 <- lm(write~read+socst) summary(lm2) plot(lm2) # plotting diagnostic plots of lm2 # Plotting all in one figure par(mfrow=c(2,2)) plot(lm2)

15