'compfactor' computes the connected components of the dummy-part of the model.
compfactor(fl, WW =FALSE)
Arguments
fl: a list of factors defining the dummies
WW: logical. Use Weeks and Williams components
Returns
A factor of the same length as the factors in the input argument. It defines the connected components. E.g. nlevels(compfactor(fl))
will yield the number of connected components.
Details
If there are more than two factors and WW=FALSE, only the first two will be used.
If WW=TRUE and length(fl) > 2, the component structure will be as in "A Note on the Determination of Connectedness in an N-Way Cross Classification" by D.L. Weeks and D.R. Williams, Technometrics, vol 6 no 3, August 1964. I.e. in each component, the coefficients within each factor are comparable, that is, their difference is estimable even if there are more than two factors. That is, one may use one reference in each factor in each component, and interpret the coefficients within a component as usual. This is not an exhaustion of all the estimable functions. There is somewhat more about this in one of the vignettes.
Examples
## create two factorsf1 <- factor(sample(300,400, replace =TRUE))f2 <- factor(sample(300,400, replace =TRUE))## find the componentscf <- compfactor(list(f1 = f1, f2 = f2))## show the third largest componentfr <- data.frame(f1, f2, cf)fr[cf ==3,]