Bar Plots

Description

Create bar plots for one or two factors scaled by frequency or precentages. In the case of two factors, the bars can be divided (stacked) or plotted in parallel (side-by-side). This function is a front end to barplot in the graphics package.

Usage

Barplot(x, by, scale = c("frequency", "percent"), conditional=TRUE,
  style = c("divided", "parallel"), 
  col=if (missing(by)) "gray" else rainbow_hcl(length(levels(by))),
  xlab = deparse(substitute(x)), legend.title = deparse(substitute(by)), 
  ylab = scale, main=NULL, legend.pos = "above", label.bars=FALSE, ...)

Arguments

Value

Invisibly returns the horizontal coordinates of the centers of the bars.

Author(s)

John Fox jfox@mcmaster.ca

See Also

barplot, legend, rainbow_hcl

Examples

with(Mroz, {
  Barplot(wc)
  Barplot(wc, col="lightblue", label.bars=TRUE)
  Barplot(wc, by=hc)
  Barplot(wc, by=hc, scale="percent", label.bars=TRUE)
  Barplot(wc, by=hc, style="parallel", 
    scale="percent", legend.pos="center")
})

Confidence Intervals by the Delta Method

Description

DeltaMethod is a wrapper for the deltaMethod function in the car package. It computes the asymptotic standard error of an arbitrary, usually nonlinear, function of model coefficients, which are named b0 (if there is an intercept in the model), b1, b2, etc., and based on the standard error, a confidence interval based on the normal distribution.

Usage

DeltaMethod(model, g, level = 0.95)
## S3 method for class 'DeltaMethod'
print(x, ...)

Arguments

Value

DeltaMethod returns an objects of class "DeltaMethod", for which a print method is provided.

Author(s)

John Fox jfox@mcmaster.ca

See Also

deltaMethod function in the car package

Examples

if (require(car)){
  DeltaMethod(lm(prestige ~ income + education, data=Duncan), "b1/b2")
  }

Dot Plots

Description

Dot plot of numeric variable, either using raw values or binned, optionally classified by a factor. Dot plots are useful for visualizing the distribution of a numeric variable in a small data set.

Usage

Dotplot(x, by, bin = FALSE, breaks, xlim, 
    xlab = deparse(substitute(x)))

Arguments

Details

If the by argument is specified, then one dot plot is produced for each level of by; these are arranged vertically and all use the same scale for x. An attempt is made to adjust the size of the dots to the space available without making them too big.

Value

Returns NULL invisibly.

Author(s)

John Fox jfox@mcmaster.ca

See Also

hist

Examples

if (require(car)){
  data(Duncan)
  with(Duncan, {
    Dotplot(education)
    Dotplot(education, bin=TRUE)
    Dotplot(education, by=type)
    Dotplot(education, by=type, bin=TRUE)
  })
}

The Gumbel Distribution

Description

Density, distribution function, quantile function and random generation for the Gumbel distribution with specified location and scale parameters.

Usage

dgumbel(x, location = 0, scale = 1)
pgumbel(q, location=0, scale=1, lower.tail=TRUE)
qgumbel(p, location=0, scale=1, lower.tail=TRUE)
rgumbel(n, location=0, scale=1)

Arguments

Author(s)

John Fox jfox@mcmaster.ca

References

See https://en.wikipedia.org/wiki/Gumbel_distribution for details of the Gumbel distribution.

Examples

x <- 100 + 5*c(-Inf, -1, 0, 1, 2, 3, Inf, NA)
dgumbel(x, 100, 5)
pgumbel(x, 100, 5)

p <- c(0, .25, .5, .75, 1, NA)
qgumbel(p, 100, 5)

summary(rgumbel(1e5, 100, 5))

Plot a Histogram

Description

This function is a wrapper for the hist function in the base package, permitting percentage scaling of the vertical axis in addition to frequency and density scaling.

Usage

Hist(x, groups, scale=c("frequency", "percent", "density"), xlab=deparse(substitute(x)), 
    ylab=scale, main="", breaks = "Sturges", ...)

Arguments

Value

This function is primarily called for its side effect — plotting a histogram or histograms — but it also invisibly returns an object of class hist or a list of hist objects.

Author(s)

John Fox jfox@mcmaster.ca

See Also

hist

Examples

    data(Prestige, package="car")
    Hist(Prestige$income, scale="percent")
    with(Prestige, Hist(income, groups=type))
    

K-Means Clustering Using Multiple Random Seeds

Description

Finds a number of k-means clusting solutions using R's kmeans function, and selects as the final solution the one that has the minimum total within-cluster sum of squared distances.

Usage

KMeans(x, centers, iter.max=10, num.seeds=10)

Arguments

Value

A list with components:

Author(s)

Dan Putler

See Also

kmeans

Examples

  data(USArrests)
  KMeans(USArrests, centers=3, iter.max=5, num.seeds=5)

Append a Cluster Membership Variable to a Dataframe

Description

Correctly creates a cluster membership variable that can be attached to a dataframe when only a subset of the observations in that dataframe were used to create the clustering solution. NAs are assigned to the observations of the original dataframe not used in creating the clustering solution.

Usage

assignCluster(clusterData, origData, clusterVec)

Arguments

Value

A factor (with integer labels) that indicate the cluster assignment for each observation, with an NA value given to observations not used in the clustering solution.

Author(s)

Dan Putler

See Also

hclust, cutree, kmeans, KMeans

Examples

  data(USArrests)
  USArrkm3 <- KMeans(USArrests[USArrests$UrbanPop<66, ], centers=3)
  assignCluster(USArrests[USArrests$UrbanPop<66, ], USArrests, USArrkm3$cluster)

Bin a Numeric Varisible

Description

Create a factor dissecting the range of a numeric variable into bins of equal width, (roughly) equal frequency, or at "natural" cut points. The cut function is used to create the factor. bin.var is a synomym for binVariable, retained for backwards compatibility.

Usage

binVariable(x, bins = 4, method = c("intervals", "proportions", "natural"),
    labels = FALSE)
    
bin.var(...)

Arguments

Value

A factor.

Author(s)

Dan Putler, slightly modified by John Fox jfox@mcmaster.ca with the original author's permission.

See Also

cut, kmeans.

Examples

summary(binVariable(rnorm(100), method="prop", labels=letters[1:4]))

Binned Frequency Distributions of Numeric Variables

Description

Bins a numeric variable, as for a histogram, and reports the count and percentage in each bin. The computations are done by the hist function, but no histogram is drawn. If supplied a numeric matrix or data frame, the distribution of each column is printed.

Usage

binnedCounts(x, breaks="Sturges", round.percents=2, 
  name=deparse(substitute(x)))

Arguments

Value

For a numeric vector, invisibly returns the vector of counts, named with the end-points of the corresponding bins. For a matrix or data frame, invisibly returns NULL

Author(s)

John Fox jfox@mcmaster.ca

See Also

hist, discreteCounts

Examples

with(Prestige, binnedCounts(income))
binnedCounts(Prestige[, 1:4])

Row, Column, and Total Percentage Tables

Description

Percentage a matrix or higher-dimensional array of frequency counts by rows, columns, or total frequency.

Usage

colPercents(tab, digits=1)
rowPercents(tab, digits=1)
totPercents(tab, digits=1)

Arguments

Value

Returns an array of the same size and shape as tab percentaged by rows or columns, plus rows or columns of totals and counts, or by the table total.

Author(s)

John Fox jfox@mcmaster.ca

Examples

if (require(car)){
    data(Mroz) # from car package
    cat("\n\n column percents:\n")
    print(colPercents(xtabs(~ lfp + wc, data=Mroz)))
    cat("\n\n row percents:\n")
    print(rowPercents(xtabs(~ hc + lfp, data=Mroz)))
    cat("\n\n total percents:\n")
    print(totPercents(xtabs(~ hc + wc, data=Mroz)))
    cat("\n\n three-way table, column percents:\n")
    print(colPercents(xtabs(~ lfp + wc + hc, data=Mroz)))
}

Frequency Distributions of Numeric Variables

Description

Computes the frequency and percentage distribution of a descrete numeric variable or the distributions of the variables in a numeric matrix or data frame.

Usage

discreteCounts(x, round.percents=2, name=deparse(substitute(x)),
  max.values=min(round(2*sqrt(length(x))), round(10*log10(length(x))), 100))

Arguments

Value

For a numeric vector, invisibly returns the table of counts. For a matrix or data frame, invisibly returns NULL

Author(s)

John Fox jfox@mcmaster.ca

See Also

binnedCounts

Examples

set.seed(12345) # for reproducibility
discreteCounts(data.frame(x=rpois(51, 2), y=rpois(51, 10)))

Plot Distribution of Discrete Numeric Variable

Description

Plot the distribution of a discrete numeric variable, optionally classified by a factor.

Usage

discretePlot(x, by, scale = c("frequency", "percent"), 
    xlab = deparse(substitute(x)), ylab = scale, main = "",
    xlim=NULL, ylim=NULL, ...)

Arguments

.

Details

If the by argument is specified, then one plot is produced for each level of by; these are arranged vertically and all use the same scale for the horizontal and vertical axes.

Value

Returns NULL invisibly.

Author(s)

John Fox jfox@mcmaster.ca

See Also

Hist, link{Dotplot}.

Examples

if (require(datasets)){
  data(mtcars)
  mtcars$cyl <- factor(mtcars$cyl)
  with(mtcars, {
    discretePlot(carb)
    discretePlot(carb, scale="percent")
    discretePlot(carb, by=cyl)
  })
}

Index Plots

Description

Index plots with point identification.

Usage

indexplot(x, groups, labels = seq_along(x), id.method = "y", type = "h", 
    id.n = 0, ylab, legend="topright", title, col=palette(), ...) 

Arguments

Value

Returns labelled indices of identified points or (invisibly) NULL if no points are identified or if there are multiple variables with some missing data.

Author(s)

John Fox jfox@mcmaster.ca

See Also

showLabels, plot.default

Examples

if (require("car")){
    with(Prestige, indexplot(income, id.n=2, labels=rownames(Prestige)))
    indexplot(Prestige[, c("income", "education", "prestige")], 
        groups = Prestige$type, id.n=2)
}

Plot a one or more lines.

Description

This function plots lines for one or more variables against another variable — typically time series against time.

Usage

lineplot(x, ..., legend)

Arguments

Value

Produces a plot; returns NULL invisibly.

Author(s)

John Fox jfox@mcmaster.ca

Examples

    if (require("car")){
    data(Bfox)
    Bfox$time <- as.numeric(rownames(Bfox))
    with(Bfox, lineplot(time, menwage, womwage))
}

Function to Merge Rows of Two Data Frames.

Description

This function merges two data frames by combining their rows.

Usage

mergeRows(X, Y, common.only = FALSE, ...)

## S3 method for class 'data.frame'
mergeRows(X, Y, common.only = FALSE, ...)

Arguments

Value

A data frame containing the rows from both input data frames.

Author(s)

John Fox

See Also

For column merges and more complex merges, see merge.

Examples

if (require(car)){
    data(Duncan)
    D1 <- Duncan[1:20,]
    D2 <- Duncan[21:45,]
    D <- mergeRows(D1, D2)
    print(D)
    dim(D)
}

Normality Tests

Description

Perform one of several tests of normality, either for a variable or for a variable by groups. The normalityTest function uses the shapiro.test function or one of several functions in the nortest package. If tests are done by groups, then adjusted p-values, computed by the Holm method, are also reported (see p.adjust).

Usage

normalityTest(x, ...)

## S3 method for class 'formula'
normalityTest(formula, test, data, ...)

## Default S3 method:
normalityTest(x, 
    test=c("shapiro.test", "ad.test", "cvm.test", "lillie.test", 
        "pearson.test", "sf.test"),
    groups, vname, gname, ...)

Arguments

Value

If testing by groups, the function invisibly returns NULL; otherwise it returns an object of class "htest", which normally would be printed.

Author(s)

John Fox jfox@mcmaster.ca

See Also

shapiro.test, ad.test, cvm.test, lillie.test, pearson.test, sf.test.

Examples

  data(Prestige, package="car")
  with(Prestige, normalityTest(income))
  normalityTest(income ~ type, data=Prestige, test="ad.test")
  normalityTest(~income, data=Prestige, test="pearson.test", n.classes=5)

Summary Statistics for Numeric Variables

Description

numSummary creates neatly formatted tables of means, standard deviations, coefficients of variation, skewness, kurtosis, and quantiles of numeric variables. CV computes the coefficient of variation.

Usage

numSummary(data, 
           statistics=c("mean", "sd", "se(mean)", "var", "CV", "IQR", 
                "quantiles", "skewness", "kurtosis"),
            type=c("2", "1", "3"),
            quantiles=c(0, .25, .5, .75, 1),
            groups)

CV(x, na.rm=TRUE)
    
## S3 method for class 'numSummary'
print(x, ...)

Arguments

Value

numSummary returns an object of class "numSummary" containing the table of statistics to be reported along with information on missing data, if there are any. CV returns the coefficient(s) of variation.

Author(s)

John Fox jfox@mcmaster.ca

See Also

mean, sd, quantile, skewness, kurtosis.

Examples

if (require("carData")){
    data(Prestige)
    Prestige[1, "income"] <- NA
    print(numSummary(Prestige[,c("income", "education")], 
    	statistics=c("mean", "sd", "quantiles", "cv", "skewness", "kurtosis")))
    print(numSummary(Prestige[,c("income", "education")], groups=Prestige$type))
    remove(Prestige)
}

Partial Correlations

Description

Computes a matrix of partial correlations between each pair of variables controlling for the others.

Usage

partial.cor(X, tests=FALSE, use=c("complete.obs", "pairwise.complete.obs"))

Arguments

Value

Returns the matrix of partial correlations, optionally with adjusted and unadjusted p-values.

Author(s)

John Fox jfox@mcmaster.ca

See Also

cor

Examples

    data(DavisThin, package="car")
    partial.cor(DavisThin)
    partial.cor(DavisThin, tests=TRUE)

Draw a Piechart With Percents or Counts in the Labels

Description

piechart is a front-end to the standard R pie function, with the capability of adding percents or counts to the pie-segment labels.

Usage

piechart(x, scale = c("percent", "frequency", "none"), 
  col = rainbow_hcl(nlevels(x)), ...)

Arguments

Author(s)

John Fox jfox@mcmaster.ca

See Also

pie, rainbow_hcl

Examples

with(Duncan, piechart(type))

Plot Bootstrap Distributions

Description

The function takes an object of class "boot" and creates an array of density estimates for the bootstrap distributions of the parameters.

Usage

plotBoot(object, confint=NULL, ...)
## S3 method for class 'boot'
plotBoot(object, confint=NULL, ...) 

Arguments

Details

Creates an array of adaptive kernal density plots, using densityPlot in the car package, showing the bootstrap distribution, point estimate ,and (optionally) confidence limits for each parameter.

Value

Invisibly returns the object produced by densityPlot.

Author(s)

John Fox jfox@mcmaster.ca

See Also

densityPlot

Examples

## Not run: 
plotBoot(Boot(lm(prestige ~ income + education + type, data=Duncan)))

## End(Not run)

Plot a probability density, mass, or distribution function.

Description

This function plots a probability density, mass, or distribution function, adapting the form of the plot as appropriate.

Usage

plotDistr(x, p, discrete=FALSE, cdf=FALSE, 
  regions=NULL, col="gray", 
  legend=TRUE, legend.pos="topright", ...)

Arguments

Value

Produces a plot; returns NULL invisibly.

Author(s)

John Fox jfox@mcmaster.ca

Examples

x <- seq(-4, 4, length=100)
plotDistr(x, dnorm(x), xlab="Z", ylab="p(z)", main="Standard Normal Density")
plotDistr(x, dnorm(x), xlab="Z", ylab="p(z)", main="Standard Normal Density",
    region=list(c(1.96, Inf), c(-Inf, -1.96)), col=c("red", "blue"))
plotDistr(x, dnorm(x), xlab="Z", ylab="p(z)", main="Standard Normal Density",
    region=list(c(qnorm(0), qnorm(.025)), c(qnorm(.975), qnorm(1)))) # same 

x <- 0:10
plotDistr(x, pbinom(x, 10, 0.5), xlab="successes", 
    discrete=TRUE, cdf=TRUE,
    main="Binomial Distribution Function, p=0.5, n=10")

Plot Means for One or Two-Way Layout

Description

Plots cell means for a numeric variable in each category of a factor or in each combination of categories of two factors, optionally along with error bars based on cell standard errors or standard deviations.

Usage

plotMeans(response, factor1, factor2, 
    error.bars = c("se", "sd", "conf.int", "none"),
    level=0.95, xlab=deparse(substitute(factor1)), 
    ylab=paste("mean of", deparse(substitute(response))),
    legend.lab=deparse(substitute(factor2)), 
    legend.pos=c("farright", "bottomright", "bottom", "bottomleft", 
                 "left", "topleft", "top", "topright", "right", "center"),
    main="Plot of Means",
    pch=1:n.levs.2, lty=1:n.levs.2, col=palette(), connect=TRUE, ...)

Arguments

Value

The function invisibly returns NULL.

Author(s)

John Fox jfox@mcmaster.ca

See Also

interaction.plot

Examples

if (require(car)){
    data(Moore)
    with(Moore, plotMeans(conformity, fcategory, partner.status, ylim=c(0, 25)))
}

Compute Pearson or Spearman Correlations with p-Values

Description

This function uses the rcorr function in the Hmisc package to compute matrices of Pearson or Spearman correlations along with the pairwise p-values among the correlations. The p-values are corrected for multiple inference using Holm's method (see p.adjust). Observations are filtered for missing data, and only complete observations are used.

Usage

rcorr.adjust(x, type = c("pearson", "spearman"), 
	use=c("complete.obs", "pairwise.complete.obs"))

## S3 method for class 'rcorr.adjust'
print(x, ...)

Arguments

Value

Returns an object of class "rcorr.adjust", which is normally just printed.

Author(s)

John Fox, adapting code from Robert A. Muenchen.

See Also

rcorr, p.adjust.

Examples

if (require(car)){
    data(Mroz)
    print(rcorr.adjust(Mroz[,c("k5", "k618", "age", "lwg", "inc")]))
    print(rcorr.adjust(Mroz[,c("k5", "k618", "age", "lwg", "inc")], type="spearman"))
    }

Read a SAS b7dat Data Set

Description

readSAS reads a SAS “b7dat” data set, stored in a file of type .sas7bdat, into an R data frame; it provides a front end to the read_sas function in the haven package.

Usage

readSAS(file, rownames=FALSE, stringsAsFactors=FALSE)

Arguments

Value

a data frame

Author(s)

John Fox jfox@mcmaster.ca

See Also

read_sas

Read an SPSS Data Set

Description

readSPSS reads an SPSS data set, stored in a file of type .sav or .por, into an R data frame; it provides a front end to the read_spss function in the haven package and the read.spss function in the foreign package.

Usage

readSPSS(file, rownames=FALSE, stringsAsFactors=FALSE,
  tolower=TRUE, use.value.labels=TRUE, use.haven=!por)

Arguments

Value

a data frame

Author(s)

John Fox jfox@mcmaster.ca

See Also

read_spss, read.spss

Read a Stata Data Set

Description

readStata reads a Stata data set, stored in a file of type .dta, into an R data frame; it provides a front end to the read.dta13 function in the readstata13 package.

Usage

readStata(file, rownames=FALSE, stringsAsFactors=FALSE,  convert.dates=TRUE)

Arguments

Value

a data frame

Author(s)

John Fox jfox@mcmaster.ca

See Also

read.dta13

Read an Excel File

Description

readXL reads an Excel file, either of type .xls or .xlsx into an R data frame; it provides a front end to the read_excel function in the readxl package. excel_sheets is re-exported from the readxl package and reports the names of spreadsheets in an Excel file.

Usage

readXL(file, rownames = FALSE, header = TRUE, na = "", sheet = 1, 
  stringsAsFactors = FALSE)
  
excel_sheets(path)

Arguments

Value

a data frame

Author(s)

John Fox jfox@mcmaster.ca

See Also

read_excel, excel_sheets

Reliability of a Composite Scale

Description

Calculates Cronbach's alpha and standardized alpha (lower bounds on reliability) for a composite (summated-rating) scale. Standardized alpha is for the sum of the standardized items. In addition, the function calculates alpha and standardized alpha for the scale with each item deleted in turn, and computes the correlation between each item and the sum of the other items.

Usage

reliability(S)

## S3 method for class 'reliability'
print(x, digits=4, ...)

Arguments

Value

an object of class reliability, which normally would be printed.

Author(s)

John Fox jfox@mcmaster.ca

References

N. Cliff (1986) Psychological testing theory. Pp. 343–349 in S. Kotz and N. Johnson, eds., Encyclopedia of Statistical Sciences, Vol. 7. Wiley.

See Also

cov

Examples

if (require(car)){
    data(DavisThin)
    reliability(cov(DavisThin))
}

Plot Means for Repeated-Measures ANOVA Designs

Description

Creates a means plot for a repeated-measures ANOVA design with one or two within-subjects factor and zero or more between-subjects factors, for data in "wide" format.

Usage

repeatedMeasuresPlot(data, within, within.names, within.levels, 
    between.names = NULL, response.name = "score", trace, xvar, 
    pch=15:25, lty=1:6, col=palette()[-1], 
    plot.means=TRUE, print.tables = FALSE)

Arguments

Value

A "trellis" object, which normally is just "printed" (i.e., plotted).

Author(s)

John Fox jfox@mcmaster.ca

See Also

Anova, OBrienKaiser

Examples

if (require(carData)){
repeatedMeasuresPlot(
  data=OBrienKaiser, 
  within=c("pre.1", "pre.2", "pre.3", "pre.4", "pre.5",
          "post.1", "post.2", "post.3", "post.4", "post.5",
          "fup.1", "fup.2", "fup.3", "fup.4", "fup.5"), 
  within.names=c("phase", "hour"), 
  within.levels=list(phase=c("pre", "post", "fup"),
                    hour = c("1", "2", "3", "4", "5")),
  between.names=c("gender", "treatment"),
  response.name="improvement",
  print.tables=TRUE)
}

if (require(carData)){
repeatedMeasuresPlot(data=OBrienKaiser, 
  within=c("pre.1", "pre.2", "pre.3", "pre.4", "pre.5",
          "post.1", "post.2", "post.3", "post.4", "post.5",
          "fup.1", "fup.2", "fup.3", "fup.4", "fup.5"), 
  within.names=c("phase", "hour"), 
  within.levels=list(phase=c("pre", "post", "fup"),
                    hour = c("1", "2", "3", "4", "5")),
  between.names=c("gender", "treatment"),
  trace="gender") # note that gender is between subjects
}

if (require(carData)){
repeatedMeasuresPlot(
  data=OBrienKaiser, 
  within=c("fup.1", "fup.2", "fup.3", "fup.4", "fup.5"), 
  within.names="hour", 
  within.levels=list(hour = c("1", "2", "3", "4", "5")),
  between.names=c("treatment", "gender"),
  response.name="improvement")
}

Reshape Repeated-Measures Data from Long to Wide Format

Description

A simple front-end to the standard R reshape function. The data are assumed to be in "long" format, with several rows for each subject.

Usage

reshapeL2W(data, within, id, varying, ignore)

Arguments

Details

Between-subjects variables don't vary by occasions for each subject. Variables that aren't listed explicitly in the arguments to the function are assumed to be between-subjects variables, and a warning is printed if their values aren't invariant for each subject (see the ignore argument).

Within-subjects factors vary by occasions for each subject, and it is assumed that the within-subjects design is regular, completely crossed, and balanced, so that the same combinations of within-subjects factors are observed for each subject.

Occasion-varying variables, as their name implies, (potentially) vary by occasions for each subject, and include one or more "response" variables, possibly along with occasion-varying covariates; these variables can be factors as well as numeric variables.

The data are reshaped so that there is one row per subject, with columns for the between-subjects variables, and each occasion-varying variable as multiple columns representing the combinations of levels of the within-subjects factors. The names of the columns for the occasion-varying variables are composed from the combinations of levels of the within-subjects factors and from the names of the occasion-varying variables. If a subject in the long form of the data set lacks any combination of levels of within-subjects factors, he or she is excluded (with a warning) from the wide form of the data.

Value

a data frame in "wide" format, with one row for each subject, columns representing the between subjects factors, and columns for the occasion-varying variable(s) for each combination of within-subjects factors.

Author(s)

John Fox jfox@mcmaster.ca

See Also

reshape, OBrienKaiser, OBrienKaiserLong.

Examples

if (require(carData)){
  OBW <- reshapeL2W(OBrienKaiserLong,
                    within=c("phase", "hour"), 
                    id="id", varying="score")
  brief(OBW)
  # should be the same as OBrienKaiser in the carData package:
  all.equal(OBrienKaiser, OBW, check.attributes=FALSE)
}

Reshape Repeated-Measures Data from Wide to Long Format

Description

The data are assumed to be in "wide" format, with a single row for each subject, and different columns for values of one or more repeated-measures variables classified by one or more within-subjects factors.

Usage

reshapeW2L(data, within, levels, varying, ignore, id = "id")

Arguments

Details

Between-subjects variables don't vary by occasions for each subject. Variables that aren't listed explicitly in the arguments to the function are assumed to be between-subjects variables. The values of these variables are duplicated in each row pertaining to a given subject.

Within-subjects factors vary by occasions for each subject, and it is assumed that the within-subjects design is regular, completely crossed, and balanced, so that the same combinations of within-subjects factors are observed for each subject. There are typically one or two within-subject factors.

Occasion-varying variables, as their name implies, (potentially) vary by occasions for each subject, and include one or more "response" variables, possibly along with occasion-varying covariates; these variables can be factors as well as numeric variables. Each occasion-varying variable is encoded in multiple columns of the wide form of the data and in a single column in the long form. There is typically one occasion-varying variable, a response variable.

There is one value of each occasion-varying variable for each combination of levels of the within-subjects factors. Thus, the number of variables in the wide data for each occasion-varying variable must be equal to the product of levels of the within-subjects factors, with the levels of the within-subjects factors varying most quickly from right to left in the within argument.

Value

a data frame in "long" format, with multiple rows for each subject (equal to the number of combinations of levels of the within-subject factors) and one column for each between-subjects and occasion-varying variable.

Author(s)

John Fox jfox@mcmaster.ca

See Also

reshapeL2W, reshape, OBrienKaiser, OBrienKaiserLong.

Examples

if (require(carData)){
  OBrienKaiserL <- reshapeW2L(OBrienKaiser, within=c("phase", "hour"), 
        levels=list(phase=c("pre", "post", "fup"), hour=1:5),
        varying=list(score=c("pre.1", "pre.2", "pre.3", "pre.4", "pre.5", 
                  "post.1", "post.2", "post.3", "post.4", "post.5", 
                  "fup.1", "fup.2", "fup.3", "fup.4", "fup.5")))
  brief(OBrienKaiserL, c(15, 15))
  m1 <- Tapply(score ~ phase + hour + treatment + gender, mean, data=OBrienKaiserL)
  m2 <- Tapply(score ~ phase + hour + treatment + gender, mean, data=OBrienKaiserLong)
  all.equal(m1, m2) # should be equal
}

if (require(carData)){
  OBrienKaiserL2 <- reshapeW2L(OBrienKaiser, within="phase", 
                  levels=list(phase=c("pre", "post", "fup")),
                  ignore=c("pre.2", "pre.3", "pre.4", "pre.5", 
                           "post.2", "post.3", "post.4", "post.5", 
                           "fup.2", "fup.3", "fup.4", "fup.5"),
                  varying=list(score=c("pre.1", "post.1", "fup.1")))
  brief(OBrienKaiserL2, c(6, 6))
  m1 <- Tapply(score ~ phase + treatment + gender, mean, data=OBrienKaiserL2)
  m2 <- Tapply(score ~ phase + treatment + gender, mean, 
               data=subset(OBrienKaiserLong, hour==1))
  all.equal(m1, m2) # should be equal
}

Stepwise Model Selection

Description

This function is a front end to the stepAIC function in the MASS package.

Usage

stepwise(mod, 
    direction = c("backward/forward", "forward/backward", "backward", "forward"), 
    criterion = c("BIC", "AIC"), ...)

Arguments

Value

The model selected by stepAIC.

Author(s)

John Fox jfox@mcmaster.ca

References

W. N. Venables and B. D. Ripley Modern Applied Statistics Statistics with S, Fourth Edition Springer, 2002.

See Also

stepAIC

Examples

# adapted from ?stepAIC in MASS
if (require(MASS)){
data(birthwt)
bwt <- with(birthwt, {
    race <- factor(race, labels = c("white", "black", "other"))
    ptd <- factor(ptl > 0)
    ftv <- factor(ftv)
    levels(ftv)[-(1:2)] <- "2+"
    data.frame(low = factor(low), age, lwt, race, smoke = (smoke > 0),
               ptd, ht = (ht > 0), ui = (ui > 0), ftv)
})
birthwt.glm <- glm(low ~ ., family = binomial, data = bwt)
print(stepwise(birthwt.glm, trace = FALSE))
print(stepwise(birthwt.glm, direction="forward/backward"))
}

Linear Model Summary with Sandwich Standard Errors

Description

summarySandwich creates a summary of a "lm" object similar to the standard one, with sandwich estimates of the coefficient standard errors in the place of the usual OLS standard errors, also modifying as a consequence the reported t-tests and p-values for the coefficients. Standard errors may be computed from a heteroscedasticity-consistent ("HC") covariance matrix for the coefficients (of several varieties), or from a heteroscedasticity-and-autocorrelation-consistent ("HAC") covariance matrix.

Usage

summarySandwich(model, ...)
    
## S3 method for class 'lm'
summarySandwich(model, 
  type=c("hc3", "hc0", "hc1", "hc2", "hc4", "hac"), ...)

Arguments

Value

an object of class "summary.lm", with sandwich standard errors substituted for the usual OLS standard errors; the omnibus F-test is similarly adjusted.

Author(s)

John Fox jfox@mcmaster.ca

See Also

hccm, vcovHAC.

Examples

    mod <- lm(prestige ~ income + education + type, data=Prestige)
    summary(mod)
    summarySandwich(mod)