df's in lme function

[Karin Groothuis-Oudshoorn <twentestat@yahoo.com>]

Hello,

I have a question about the degrees of freedom in the lme function of S-plus. I have data with 2 levels (subjects and repeated measurements of the subjects). The number of subjects is 94. Furthermore I have 3 covariates (say var1, var2, var3) that are level one variables. And one covariate is at the second level (bmi=body mass index). The total number of observations is 289. The outcome variable is denoted by outcome.

The fit I requested for is as follows:

fit_lme(fixed=outcome~bmi+var1+var2+var3,random=~1|subject,data="">

The summary of the fitobject of lme gives for the covariate of the second level 92 as expected, as degrees of freedom. For the other covariates, that are level one variables it gives 192 as degrees of freedom. However I expected not 192 as df's for each variable, but 289-4 (=total number of variables exept for the intercept) - 1 = 284.

Can anyone give me an explanation of the given number of degrees of reedom? Does it have to do with the intercept and the fact that for each subject an separate intercept is estimated? Does there exist different formula to use in the multilevel-world? THe formula I used was taken from the book of Snijders & Bosker.

Thanks in advance,

Karin Oudshoorn

---

Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now