For continuous
covariates it is a common to fit them as linear terms. However, one sometime
finds that they are first used to define categories. For example, various age
groups might be defined using age. If these are fitted then it is usual to fit
them as a factor. The levels of the factor are then implicitly taken as
unordered.

Similarly,
for directly defined ordered covariates it is common to fit the covariate as an
unordered factor.

A more
rational approach might be to define orthogonal contrasts on the covariate,
gradually fitting higher terms. If the covariate was continuous, then another
possibility, less extreme than simply using a liner predictor would be to fit a
spline.

In this
project various strategies for fitting ordered categorical covariates will be
fitted with a view to investigating possible biases and gains or losses in
efficiencies.