# lessRstats.com/lessR.r # tour of lessR functions for data analysis # Purpose: Provide basic statistical computations for the analyses presented # in intro stat texts and more # Get R # http://r-project.org # one time only, to get the lessR functions onto your computer # if asked to install into a personal library, say Yes # install.packages("lessR") # Begin each R session by loading the lessR functions library(lessR) # help # ------------------------------------------------------------------------ Help() # list the lessR help values by subject Help(lessR) # access to the full lessR manual and News items Help(Read) # read data from one of many different formats into a data table called mydata # with the same Read statement: csv, tab-delimited text, Excel, SPSS, and SAS # ------------------------------------------------------------------------ # Categorical: Gender coded M, F; Dept coded ACCT, ADMN, FINC, MKTG, SALE # Numeric: Salary and Years #mydata <- Read() # or rd(), browse for text, Excel, SPSS, SAS, or R data file #head(mydata) # R function, see variable names and first 6 rows of data #mydata <- Read("http://lessRstats.com/data/employee.csv") # read data from web #mydata <- Read("http://lessRstats.com/data/employee.csv", row.names=1) # row names # data included as part of lessR mydata <- rd("Employee", format="lessR") # this example includes variable labels details() # provides full details of the data frame, defaults to mydata db() # brief version, same as provided by Read # can also read a text file o fixed width formatted data # example provided later # distribution of a categorical variable # ------------------------------------------------------------------------ # all the same BarChart(Dept) # or bc(Dept) bc(Dept) # all function calls have a 2 or 3 digit abbreviation BarChart(Dept, data=mydata) # mydata is the default data frame (table) bc(Dept, fill="colors") # specify a more colorful display bc(Dept, rotate.x=45, offset=1) # value labels rotated and offset # other summaries PieChart(Dept) # or pc(Dept), provides a doughnut or ring chart PieChart(Dept, hole=0) # full pie chart Plot(Dept) # or Plot(Dept), 1-categorical variable bubble plot SummaryStats(Dept) # or ss(Dept), no graphics # read Mach IV data, integer Likert responses from 0 to 5 mydata <- Read("Mach4", format="lessR") # read a data file included with lessR # form a Bubble Plot Frequency Matrix (BPFM) from a range of x-variables # each line is a bubble plot of frequencies for a single variable Plot(c(m06,m07,m09,m10)) # for each bubble, lighten fill color, make border black Plot(m06:m12, trans=.8, coloe="gray50") # create BPFM for entire Mach IV scale and store as a pdf file LikertCats <- c("Strongly Disagree", "Disagree", "Slightly Disagree", "Slightly Agree", "Agree", "Strongly Agree") Plot(m01:m20, value.labels=LikertCats, pdf.file="MachFreqs.pdf") Plot(m06) # Integrated Violin/Box/Scatterplot (VBS plot) # back to the Employee data mydata <- rd("Employee", format="lessR") # includes variable labels # distribution of a numeric (continuous) variable # ------------------------------------------------------------------------ Plot(Salary) # Integrated Violin/Box/Scatterplot (VBS plot) Plot(Salary, by1=Gender) # Trellis VBS plots for Gender Histogram(Salary) # or hs(Salary) hs(Salary, bin.start=30000, bin.width=12000, bin.end=150000) # common options Density(Salary) # or dn(Salary) BoxPlot(Salary) # or bx(Salary), a call to Plot with vbs.plot="b" SummaryStats(Salary) # or ss(Salary) ss.brief(Salary) # brief output hs(Salary, Rmd="Salary") # generate an R markdown file of all the methods #hs(Salry) # ERROR, misspelled variable name to show lessR error messages # for ordered values, such as by time (just for illustration here) LineChart(Salary) # or lc(Salary) Plot(Salary, run=TRUE) Plot(Salary, run=TRUE, center.line="off", show.runs=TRUE) Plot(Salary, run=TRUE, lwd=0) # points only Plot(Salary, run=TRUE, size=0) # line only Plot(Salary, run=TRUE, area=TRUE) # distributions for multiple variables (writes to pdf files) # ------------------------------------------------------------------------ # the corresponding standard R functions do not take multiple variables # specify multiple variables with c function or : Histogram(c(Salary, Years)) # c is R combine function BarChart(Gender:Satisfaction) # : indicates a range of variables in R Histogram() # a histogram for each numeric variable in mydata data table CountAll() # or ca(), a bar chart or histogram for each variable in mydata # relationship of two variables # ------------------------------------------------------------------------ # two numeric variables yields traditional scatter plot Plot(Years, Salary) # or Plot(Years, Salary) Plot(Years, Salary, auto=TRUE) # or Plot(Years, Salary) Plot(Years, Salary, ellipse=TRUE, fit="loess") # common options Plot(Years, Salary, ellipse=c(.50,.75,.90)) # 3 data ellipses Plot(Years, Salary, size=3) # bubbles # classify by a 3rd variable, which is categorical Plot(Years, Salary, by=Gender) Plot(Years, Salary, by=Gender, color=c("green", "brown")) Plot(Years, Salary, by=Gender, color=c("darkgreen", "brown"), shape=c("F","M")) Plot(Years, Salary, by=Gender, color=c("darkgreen", "brown"), fit="ls") # categorical with a numeric variable, shows scatter about each group mean Plot(Dept, Salary) # Dept is categorical, Salary is numeric SummaryStats(Salary, by=Dept) # stats, for each level of a grouping variable ss.brief(Salary, by=Dept) # brief output # two categorical variables (factors) BarChart(Gender, by=Dept) # text from ss.brief bc(Gender, by=Dept, beside=TRUE, horiz=TRUE) # common options Plot(Gender, Dept) # scatter (bubble) plot in place of traditional bar chart SummaryStats(Gender, Dept) # 4 cross-tab tables bc(Dept, by=Gender, beside=TRUE) # the proportion of data values by fill variable within each group bc(Dept, by=Gender, proportion=TRUE) Plot(Dept, Salary) # Plot refers to what is plotted in the scatter plot as the values # Plot default for values is "data" Plot(Dept, Salary, values="mean") Plot(c(Pre, Post), Salary) Plot(c(Pre, Post), Salary, fit="ls") # Cleveland dot plot test Plot(Salary, row.names) Plot(c(Pre, Post), row.names) Plot(row.names, Salary, means=FALSE, rotate.x=60, offset=1, xlab="") Plot(Salary, row.names, fill="black", color="black", sort.y=TRUE) style(panel.fill="off", grid.color="off") Plot(Salary, row.names, sort.y=TRUE, segments.y=TRUE) style() # return to default style # set system values # ------------------------------------------------------------------------ # color themes Help(theme) style("orange", sub.theme="black") hs(Salary) Plot(Years, Salary, ellipse=TRUE, fit="loess") # common options style("darkred") hs(Salary) style("gray") hs(Salary) Plot(Years, Salary, ellipse=TRUE, fit="loess") # common options # more system settings, can also be set by individual function calls style(brief=TRUE) # partial text output style(quiet=TRUE) # no text output style(n.cat=6) # treat integer variables with <=6 unique values as categorical style(brief=FALSE) # back to default style(quiet=FALSE) style(n.cat=0) # analysis of a mean # ------------------------------------------------------------------------ ttest(Salary) # or tt(Salary) for confidence interval tt.brief(Salary, mu0=70000) # brief output, for CI and hypothesis test tt.brief(n=37, m=63795.557, s=21799.533) # input summary stats instead of data ttestPower(n=20, s=5) # compare 2 groups, with graphics # ------------------------------------------------------------------------ ttest(Salary ~ Gender) # or tt(Salary ~ Gender), compare means for two groups #tt(Gender ~ Salary) # ERROR, do it wrong to illustrate lessR error messages tt.brief(Salary ~ Gender) # brief output tt(n1=18, n2=19, m1=71147.458, m2=56830.598, s1=23128.436, s2=18438.456) # from stats ttestPower(n1=14, n2=27, s1=4, s2=6, msmd=.5) # Pre and Post are variables that are two matched sets of data values ttest(Pre, Post) # t-test to compare group means, data entered as two vectors ttest(Pre, Post, paired=TRUE) # t-test of paired differences # analysis of variance, compare 2 or more groups # ------------------------------------------------------------------------ # warpbreaks is a data set contained in R ANOVA(breaks ~ tension, data=warpbreaks) # or av, one-way ANOVA # two-factor between-groups ANOVA with replications and interaction ANOVA(breaks ~ wool * tension, data=warpbreaks) # randomized blocks design with the second term the blocking factor ANOVA(breaks ~ wool + tension, data=warpbreaks) # regression analysis # ------------------------------------------------------------------------ # single predictor variable regression reg(Salary ~ Years) # or reg, or reg.brief, equivalent of Excel regression reg.brief(Salary ~ Years) # reg.brief equivalent of Excel regression #reg(Salary ~ 1) # null model #reg.brief(Salary ~ 1) # null model # multiple regression reg(Salary ~ Years + Pre + Post) # multiple reg # output to an object r <- reg(Salary ~ Years + Pre + Post) # multiple reg, save output to object r r # see all the output names(r) # see the output segments available for viewing or further analysis r$out_estimates # see just the piece of the estimates # create R markdown file that does full interpretation of the output r <- reg(Salary ~ Years + Pre + Post, Rmd="MultReg") # R markdown # nested models analysis Nest(Salary, c(Years), c(Pre, Post)) # compare reduced to full model, with common data # logit analysis, with a classification table Logit(Gender ~ Salary) # transformations # ------------------------------------------------------------------------ # use Recode to reverse score four Likert variables: m01, m02, m03, m10 mydata <- Read("Mach4", format="lessR") # read a data file included with lessR mydata <- Recode(c(m01:m03,m10), old=0:5, new=5:0) # R has no recode function mydata <- rd("Employee", format="lessR", quiet=TRUE) # internal data, no console output mydata <- Transform(Salary = Salary / 1000) # transform by formula males <- Subset(Gender=="M", columns=c(Years, Salary)) # only Males, two vars mydata <- Sort(Salary) # default from smallest to largest mydata <- Sort(Salary, direction="-") # can also specify directiont mydata <- Sort(row.names) # unlike standard R, can sort by row name # also Merge for vertical or horizontal merge of two data tables # variable labels # ------------------------------------------------------------------------ # read data file w/o variable labels, and then separately read labels mydata <- Read("http://lessRstats.com/data/employee.csv") mylabels <- VariableLabels("http://lessRstats.com/data/employee_lbl.csv") # read data file w/o variable labels, and then read labels from console mydata <- Read("http://lessRstats.com/data/employee.csv") lbl <- " Years, time of company employment Gender, Male or Female Dept, department employed Salary, annual salary Satisfaction, satisfaction with work environment HealthPlan, 1=GoodHealth 2=YellowCross 3=BestCare Pre, Test score on legal issues before instruction Post, Test score on legal issues after instruction " mylabels <- VariableLabels(lbl) # modify/display a single variable label mylabels <- VariableLabels(Salary, "Annual Salary (USD)") # add/modify 1 label VariableLabels(Salary) # list the contents of a single variable label # display all variable labels db() # also the variable names and sample values vl() # write a data table to an external file # ------------------------------------------------------------------------ Write("myfile") # write default mydata data table to myfile in csv format Write("myfile", row.names=FALSE) # preferred if row names are only row numbers Write("myfile", format="Excel", row.names=FALSE) # to Excel data table # after reading original data and doing the transformations, save as is Write("myfile", format="R") # native R format, straight copy of mydata # correlation matrices and factor analysis # ------------------------------------------------------------------------ mydata <- Read("Mach4", format="lessR") # read a data file included with lessR # calculate the correlations and store in mycor head(mydata) mydata <- Recode(c(m03, m04, m06, m07, m09:m11, m14, m16, m17, m19), old=0:5, new=5:0) mycor <- cr(m01:m20) efa(n.factors=4) # confirmatory factor analysis of 4-factor solution of Mach IV scale # Hunter, Gerbing and Boster (1982) MeasModel <- " Deceit =~ m07 + m06 + m10 + m09 Trust =~ m12 + m05 + m13 + m01 Cynicism =~ m11 + m16 + m04 Flattery =~ m15 + m02 " c <- cfa(MeasModel) # view all the output c # names of each output segment names(c) # view just the scale reliabilities c$out_reliability # correlation matrix operations mycor <- cr(m01:m20, graphics=TRUE) mydata <- corReorder() # simple re-order algorithm by highest remaining cor prop() scree() # read fixed width data, requires parameters: widths, color.names # ------------------------------------------------------------------------ rep(1,20) # standard R function, here a 1 listed 20 times to("m",20) # lessR, names sequential vars with same widths [unlike paste0("m", 1:20)] # specify the width of each data column, then match with the column name # the Mach4 scale is 6-pt Likert data, so each response is a single column mydata <- Read("http://lessRstats.com/data/Mach4Plus.fwd", widths=c(4,2,1,5,1,rep(1,20)), col.names=c("ID", "Age", "Gender", "Code", "Form", to("m",20)))