Relevel: Reorder and combine levels of a factor

Description

The levels of a factor are re-ordered so that the levels specified by ref appear first and remaining levels are moved down. This is useful for contr.treatment contrasts which take the first level as the reference. Factor levels may also be combined; two possibilities for specifying this are supported: hard coding or table look-up.

Usage

# S3 method for factor
Relevel( x, ref, first = TRUE, collapse="+",
                 xlevels=TRUE, nogroup=TRUE, … )

Arguments

A(n unordered) factor

ref

Vector, list or data frame, array, matrix or table.

If ref is a vector (integer or character), it is assumed it contains the names or numbers of levels to be the first ones; non mentioned levels are kept.

If ref is a list (but not a data frame), factor levels mentioned in each list element are combined. If the list is named the names are used as new factor levels, otherwise new level names are constructed from the old.

If ref is a data frame or 2-dimensional array, matrix or table, the first column is assumed to have unique levels of x and the second to have groupings of this, respectively.

first

Should the levels mentioned in ref (if it is a list) come before those not?

collapse

String used when constructing names for combined factor levels.

xlevels

Logical. Should all levels in the 2nd column of ref be maintained as levels of the result, or (if FALSE) only the actually occurring.

nogroup

Logical. Should levels present in the input but not in the 1st column of ref be maintained as levels after the grouping? If FALSE), NAs will be returned for such elements.

…

Arguments passed on to other methods.

Value

An unordered factor, where levels of x have been reordered and/or collapsed.

Details

The facility where ref is a two-column matrix mimics the SAS-facility of formats where a dataset can be used to construct a format --- SAS format is the grouping tool for variable values. Using this facility will preserve the order of levels if ref[,2] is a factor.

Examples

Run this code

# NOT RUN {
# Grouping using a list (hard coding)
#
ff <- factor( sample( letters[1:5], 100, replace=TRUE ) )
table( ff, Relevel( ff, list( AB=1:2, "Dee"=4, c(3,5) ) ) )
table( ff, Relevel( ff, list( 5:4, Z=c("c","a") ), coll="-und-",first=FALSE ) )

## Grouping using a two-column matrix as input:
## A factor with levels to be grouped together
ff <- factor( c("Bear","Bear","Crocodile","Snake","Crocodile","Bear") )
ff
## A grouping table
( gg <- data.frame( Animal=c("Bear","Whale","Crocodile","Snake","Eagle"),
                     Class=c("Mammal","Mammal","Reptile","Reptile","Bird") ) )
Relevel( ff, gg, xlevels=FALSE )
Relevel( ff, gg )
Relevel( ff, gg[c(1:5,5:1),] )

## This crashes with an error
try( Relevel( ff, rbind( gg, c("Bear","Reptile") ) ) )

ff <- factor(c(as.character(ff),"Jellyfish","Spider"))
Relevel( ff, gg )
Relevel( ff, gg, xlevels=FALSE )
# If you do not want unknown animals classified, this returns NAs:
Relevel( ff, gg, nogroup=FALSE )
# Will give error because re-classification is ambiguous
( GG <- rbind(gg,c("Eagle","Mammal"),c("Snake","Mammal")) )
try( Relevel( ff, GG ) )
# }