- A formula conforming to the conventions of
loglm, that is, it may allow dimension numbers to stand in for names when specifying a log-linear model.
The model fitting function
loglm fits log-linear models to frequency data using iterative proportional scaling. To specify the model the user must nominate the margins in the data that remain fixed under the log-linear model. It is convenient to allow the user to use dimension numbers, 1, 2, 3, ... for the first, second, third, ..., margins in a similar way to variable names. As the model formula has to be parsed by
terms, which treats
1 in a special way and requires parseable variable names, these formulae have to be modified by giving genuine names for these margin, or dimension numbers.
denumerate replaces these numbers with names of a special form, namely
n is replaced by
.vn. This allows
terms to parse the formula in the usual way.
A linear model formula like that presented, except that where dimension numbers, say
n, have been used to specify fixed margins these are replaced by names of the form
.vn which may be processed by
denumerate(~(1+2+3)^3 + a/b) ## which gives ~ (.v1 + .v2 + .v3)^3 + a/b
Documentation reproduced from package MASS, version 7.3-45. License: GPL-2 | GPL-3