[s] C.I. associated with poly.transform

["Theresa Hannah" <theresa.hannah@ualberta.ca>]

Hello,

 

I have a model (glm.nb) to which I would like to extract the coefficients. I use a 2^nd order polynomial for one of the terms, and I’ve read that I should use poly.transform to derive the more accurate coefficients transformed from the orthogonal form. However, I would like to present the 90% confidence intervals associated with the coefficients. The command ‘confint’ will do it for most terms; does anyone know how to generate 90% confidence intervals on the transformed coefficients (i.e. in conjunction with the poly.transform command)?

 

The coding I’m using in S-plus 6.2 for Windows is:

 

 

patchocc.nb4<-glm.nb(yrsocc11~cnt93+poly(age,2)+log(sw+0.01), data="" na.action="">

confint(patchocc.nb4, level=0.90)

summary(patchocc.nb4)

anova(patchocc.nb4, test="Chi")

poly.transform(poly(age,2), coef(patchocc.nb4)[c(1,3,4)])

 

Thanks in advance,

Theresa