ran.CMData {CorrBin}

Generate a random CMData object

Description

Generates random exchangeably correlated multinomial data based on a parametric distribution or using resampling. The Dirichlet-Multinomial, Logistic-Normal multinomial, and discrete mixture multinomial parametric distributions are implemented. All observations will be assigned to the same treatment group, and there is no guarantee of a specific order of the observations in the output.

Usage

ran.CMData(n, ncat, clustersize.gen, distribution)

Arguments

n

number of independent clusters to generate

ncat

number of response categories

clustersize.gen

either an integer vector specifying the sizes of the clusters, which will be recycled to achieve the target number of clusters n; or a function with one parameter that returns an integer vector of cluster sizes when the target number of clusters n is passed to it as a parameter

distribution

a list with two components: "multinom.gen" and "param" that specifies the generation process for each cluster. The "multinom.gen" component should be a function of three parameters: number of clusters, vector of cluster sizes, and parameter list, that a matrix of response counts where each row is a cluster and each column is the number of responses of a given type. The "param" component should specify the list of parameters needed by the multinom.gen function.

Value

a CMData object with randomly generated number of responses with sample sizes specified in the call

Author(s)

Aniko Szabo

See Also

CMData for details about CMData objects; multinom.gen for built-in generating functions

Examples

# Resample from the dehp dataset
data(dehp)
ran.dehp <- ran.CMData(20, 3, 10, list(multinom.gen=mg.Resample, param=list(data=dehp)))

# Dirichlet-Multinomial distribution with two treatment groups and random cluster sizes
binom.cs <- function(n){rbinom(n, p=0.3, size=10)+1}
dm1 <- ran.CMData(15, 4, binom.cs, 
                  list(multinom.gen=mg.DirMult, param=list(shape=c(2,3,2,1))))
dm2 <- ran.CMData(15, 4, binom.cs, 
                  list(multinom.gen=mg.DirMult, param=list(shape=c(1,1,4,1))))
ran.dm <- rbind(dm1, dm2)

# Logit-Normal multinomial distribution
ran.ln <- ran.CMData(13, 3, 3, 
                     list(multinom.gen=mg.LogitNorm,
                          param=list(mu=c(-1, 1), sigma=matrix(c(1,0.8,0.8,2), nr=2))))

# Mixture of two multinomial distributions
unif.cs <- function(n){sample(5:9, size=n, replace=TRUE)}
ran.mm <- ran.CMData(6, 3, unif.cs, 
                     list(multinom.gen=mg.MixMult,
                          param=list(q=c(0.8,0.2), m=rbind(c(-1,0), c(0,2)))))