allComb {opticut}

Finding All Possible Binary Partitions

Description

These functions are used to find all possible binary partitions. Finding all combinations require a classification vector with K > 1 strata.

Usage

allComb(x, collapse)
kComb(k)
checkComb(x)

Arguments

x

a vector for allComb (can be of any type but treated as factor, must have at least 2 unique values); and a numeric matrix for checkComb.

collapse

character, what to paste between levels. Defaults to getOption("ocoptions")$collapse.

k

numeric, number of levels (strata) in a given classification (K > 1).

Value

kComb returns a contrast matrix corresponding to all possible binary partitions of the factor with K levels. Complements are not counted twice, i.e. (0,0,1,1) is equivalent to (1,1,0,0). The number of such possible combinations is M = 2^(K - 1) - 1.

allComb takes a classification vector with at least 2 levels and returns a model matrix with binary partitions.

checkComb checks if combinations are unique and non-complementary (misfits are returned as attributes). Returns a logical value.

Author(s)

Peter Solymos <[e-mail]>

See Also

opticut for the user interface.

rankComb and lorenz for alternative partitioning algorithms.

Examples

kComb(k = 2)
kComb(k = 3)
kComb(k = 4)

## finding all combinations
(f <- rep(LETTERS[1:4], each=2))
(mc <- allComb(f, collapse = "_"))
## checking for complementary entries
checkComb(mc) # TRUE
## adding complementary entries to the matrix
mc2 <- cbind(z = 1 - mc[,1], mc[,c(1:ncol(mc), 1)])
colnames(mc2) <- 1:ncol(mc2)
mc2
checkComb(mc2) # FALSE