MConjoint (version 0.1)

mc.good.designs: mc.good.design

Description

given a set of m cards, find "good" designs with cards rows

Usage

mc.good.designs(orig.set, cards = NULL, slack = 1, tol = 0.2, no.replace = TRUE, size = 100, max.trials = 1e+06)

Arguments

orig.set

a design of length m

cards

The number of cards in each "good" design found.

slack

How much the number of each factor can vary in a "good" design

tol

The largest cross correlation in a "good" design

no.replace

Sample without replacement: TRUE or FALSE

size

The number of "good" designs to find

max.trials

The maximum number of designs to look at

Value

A despack with the following field filled

cards

set equal to orig.set

samps

a list of samples, the row numbers of the corresponding designs

designs

the good designs found

Details

The function takes samples with cards rows from the orig.design. For each sample it checks whether the design is "good". A design is said to be good if it is balanced (for each factor each level occurs about the same number of times, the maximum difference is slack) and the different factors are uncorrelated (maximum cross correlation is tol). Sampling continues (with or without replacement depending on no.replace) until one of size good designs are found, all designs have been checked, or max.trials designs have been checked. If fewer than size design are found then a warning is printed.

Examples

data(hire.questionaire)


#default
mc.good.designs(hire.questionaire$design)

#look for 7 card designs, with the cross correlation tolerance increased to .3

#mc.good.designs(hire.questionaire$design,7,tol=.3)