Hello
I have just begun constructing poisson regression
models to explain how moss species diversity is related to various
environmental variables.
I have some very basic questions about how S-plus works when using the
add1 function and when using nested variables (S-plus release 2, windows
98 OS).
1. I am using the add1 function in order to obtain Cp values (i'm still
not entirely sure what these are) to use as criteria for the addition of
explanatory variables while conducting manual forward selection.
It seems like the choosing the lowest cp values does not work for
categorical variables (type=factor) because they have more than one degree
of freedom as some of them have very high Cps but explain alot of the
variation if they are added to the model.
Should I be accounting for the degrees of freedom of each variable when
looking at Cp values?
2. In my model, one of my explanatory variables is nested within another:
mesosite within stand. When entering these variables in my model I use
the notation given in the S-plus help which is: meso/stand.
When I use this however, stand seems to dissapear as a variable (using the
summary or the add1 function). I am wondering what S-plus is doing with
stand. Is all of the variation between stands then acounted for by the
mesosite?
Thank-you very much to anyone who can answer either of these questions.
Suzanne
|