CLO stands for constrained linear ordination, and
is fitted with a statistical class of models called
reduced-rank vector generalized linear models
(RR-VGLMs). It allows for generalized reduced-rank regression
in that response types such as Poisson counts and presence/absence
data can be handled.
Currently in the VGAM package, rrvglm is
used to fit RR-VGLMs. However, the Author's opinion is that
linear responses to a latent variable (composite environmental
gradient) is not as common as unimodal responses, therefore
cqo is often more appropriate.
The new CLO/CQO/CAO nomenclature described in Yee (2006).