[R] creating your own glm link function

[Anne G]

In sas and matlab, I have been using a modified link
function.

I use a binomial distribution with the link log(p-.5/1-p)
instead of log(p/1-p) which is the usual logit link. This
allows the probabilities to go from .5 to 1 instead of 0 to
1. which is typical when probabilities vary between chance
(.5) and certainty (1) in psychophysics experiments.

I am trying to access if R can do it. If so could someone
show me how to set up the link function so I can call it in
the glm function?

I know very little about R, the help reads like greek to me,
and I can't find an example anywhere.

Getting desperate!

Anne