question about beta regressions

Hi Marie,

betareg uses a logit link function, so the negative intercept just means 
that the intercept proportion is less than one. To convert it back to a 
proportion, use plogis(InterceptValue).

The intercept is the logit(proportion) that you would get if all your 
covariates were set to zero and the landfill treatment is set to its 
baseline. You might get a more interpretable intercept if you centre 
your explanatory variables (see ?scale). You could also look at changing 
the default contrasts to "contr.sum" which will give comparisons to the 
grand mean for your landfill variable. See ?contr.sum. To do this, just use:

options(contrasts=c("contr.sum", "contr.poly"))

Hope this helps,

Simon.

question about beta regressions

Thread (4 messages)