icsPlot: Test of Marginal Proportion for Clustered Data

In htestClust: Reweighted Marginal Hypothesis Tests for Clustered Data

Test of Marginal Proportion for Clustered Data

Description

Function to visualize informative cluster size. Plots within-cluster summary statistic from quantitative variables against the size of each cluster. For categorical variables, a barplot of category proportions for quantiles of cluster size is produced.

Usage

icsPlot(
  x,
  id,
  FUN = c("mean", "median", "var", "sd", "range", "IQR", "prop"),
  breaks,
  xlab = NULL,
  ylab = NULL,
  legend = c(TRUE, FALSE),
  ...
)

Arguments

x	vector of data values. Alternatively a two-dimensional table or matrix.
id	a vector which identifies the clusters, with length equal to length of x; ignored if x is a matrix or table.
FUN	the name of the function that produces the desired intra-cluster summary statistic.
breaks	a single number giving the number of desired quantiles for the barplot of categorical variables with >2 categories.
xlab	a label for the x axis, defaults to "cluster size".
ylab	a label for the y axis, defaults to a description of FUN of x.
legend	a logical indicating whether a legend should be included in a barplot.
...	further arguments to be passed to or from methods.

Details

If x is a matrix or table and x has exactly two columns, the first column should contain the cluster sizes and the second column the respective intra-cluster summary statistic (e.g., mean, variance) that will be plotted against cluster size.

If x has more than two columns, the first column is assumed to contain the cluster size and the subsequent columns the counts of intra-cluster observations belonging to the different categorical variable levels. If there are exactly two categorical levels (e.g., x has exactly three columns), a scatterplot of the proportion of intracluster observations belonging to the first category will be plotted against the cluster size. If the number of categories is > 2, a barplot of category proportions against quantiles of cluster size is produced.

Standard graphical parameters can be passed to icsPlot through the ... argument.

Examples

data(screen8)
## VECTOR INPUT
## plot average math score by cluster size
icsPlot(x = screen8$math, id = screen8$sch.id, pch = 20)

## plot proportion of females by cluster size
icsPlot(screen8$gender, screen8$sch.id, pch = 20, main = "Female proportion by cluster size")

## barchart of activity proportion by quartile of cluster size
icsPlot(x = screen8$activity, id = screen8$sch.id)

## TABLE INPUT
## Plot intra-cluster variance of math score by cluster size
cl.size <- as.numeric(table(screen8$sch.id))
tab1 <- cbind(cl.size, aggregate(screen8$math, list(screen8$sch.id), var)[,2])
colnames(tab1) <- c("cl.size", "variance")
icsPlot(x = tab1, pch = 17, main = "math score variance by cluster size")

## barchart of activity proportion across five quantiles of cluster size
tab2 <- cbind(cl.size, table(screen8$sch.id, screen8$activity))
icsPlot(tab2, breaks = 5)

htestClust documentation built on May 18, 2022, 9:05 a.m.