htestClust (version 0.2.2)

icstestClust: Test for Informative Cluster Size

Description

Performs a test for informative cluster size.

Usage

icstestClust(x, id, test.method = c("TF", "TCM"), B = 1000, print.it = TRUE)

Arguments

x

a vector of numeric responses. Can also be a data frame.

id

a vector or factor object which identifies the clusters; ignored if x is a data frame. The length of id must be the same as the length of x.

test.method

character string specifying the method of construction for the test statistic. Must be one of "TF" or "TCM".

B

the number of bootstrap iterations.

print.it

a logical indicating whether to print the progression of bootstrap iterations.

Value

A list with class "htest" containing the following components:

statistic

the value of the test statistic.

p.value

the p-value of the test.

method

a character string indicating the test performed and the method of construction.

data.name

a character string giving the name(s) of the data.

Details

The null is that the marginal distributions of the responses are independent of the cluster sizes. A small p-value is evidence for the presence of informative cluster size.

When test.method = "TF", the test statistic is constructed based on differences between the null and alternative distribution functions. "TF" is the suggested method when there are a large number of unique cluster sizes and the number of clusters of each size is small. When test.method = "TCM", the test statistic is a multisample Cramer von Mises-based test. This method is recommended when there are a small number of possible cluster sizes. See Nevalainen et al. (2017) for more details.

When x is a data frame, the first column should contain values denoting cluster membership and the second column the responses.

This test is computationally intensive and can take significant time to execute. print.it defaults to TRUE to identify the bootstrap progression.

References

Nevalainen, J., Oja, H., Datta, S. (2017) Tests for informative cluster size using a novel balanced bootstrap scheme. Statistics in Medicine, 36, 2630--2640.

Examples

# NOT RUN {
  data(screen8)
  ## using vectors
  ## test if cluster size is related to math scores
  icstestClust(screen8$math, screen8$sch.id, B=100)

  ## same test, but using a data frame and supressing iterations
  tdat <- data.frame(screen8$sch.id, screen8$math)
  icstestClust(tdat, B=100, print.it = FALSE)
# }