The answer is in Pinheiro and Bates (2000), "Mixed-effects Models in S
and S-PLUS". See pages 90-91
"Bernd Puschner" <bernd.puschner@bkh-guenzburg.de> writes:
> dear s-plus users,
>
> thanks to douglas bates, bert gunter, and manuela huso for answering my
> question about the meaning of the intercept's p-value in lme. the answer is:
> The p-value is for the (marginal) test of the Intercept = 0 versus Intercept
> != 0.
>
> I have another question pertaining to lme (to which I couldn't find answers
> in pinheiro/bates: "mixed effects..."): when computed a lme-model (of change
> in time of the criterion variable) with two predictors, one continuous, the
> other factorial (3 categories):
>
> model.lme <- lme(data = iip.in, random = ~in.therm | code, fixed = oqsd ~
> (x1 + as.factor(thern)) * in.therm, na.action = na.omit),
>
> the fixed effects part of the output looks like this
>
> > summary(model.lme)
> ...
> Fixed effects: oqsd ~ (x1 + as.factor(thern)) * in.therm
> Value Std.Error DF t-value p-value
> (Intercept) 40.90453 0.7882101 932 51.89547 <.0001
> x1 -0.07581 0.2215110 532 -0.34226 0.7323
> as.factor(thern)1 -1.04852 0.8783406 532 -1.19375 0.2331
> as.factor(thern)2 0.09866 0.4917715 532 0.20062 0.8411
> in.therm -0.38881 0.0450593 932 -8.62895 <.0001
> x1:in.therm -0.00077 0.0133241 932 -0.05813 0.9537
> as.factor(thern)1in.therm 0.05369 0.0457854 932 1.17262 0.2412
> as.factor(thern)2in.therm -0.01635 0.0306450 932 -0.53351 0.5938
>
> ...
> Number of Observations: 1472
> Number of Groups: 536
>
> How exactly does S-Plus calculate the degrees of freedom?
>
> Thanks for any ideas.
--
Douglas Bates bates@stat.wisc.edu
Statistics Department 608/262-2598
University of Wisconsin - Madison http://www.stat.wisc.edu/~bates/
|