Fitted values of 'lme' function result

Question

I fitted a linear mixed model with R as follows.

lmme3cs = lme(LAZ ~ group_k + x1 + x2 + x4 + x6 + x1_k + x2_k + x4_k + x6_k,
              random = ~1|SECTORID/CHILDUID,
              correlation = corCompSymm(form = ~1|SECTORID/CHILDUID), data=dat2c)

It has 50043 observations, but the object 'lmme3cs$fitted' is a 50043 by 3 matrix which has three columns fixed, SECTORID, and CHILDUID. What are these three columns?

Answer 1

They are the fitted values at different levels of grouping. For example, if you fit

> library(nlme)
> fmOxide <- lme(Thickness ~ 1, Oxide, ~1|Lot/Wafer)

you can get the predicted overall outcome by

> fitted(fmOxide, level = 0)

and it is 2000.153. The predicted outcomes for each Lot are:

> fitted(fmOxide, level = 1)

and there are eight outocomes as there are eight Lots. The predicted outcomes for each Wafer, which are nested within Lots, are:

> fitted(fmOxide, level = 2)

If you run

> fitted(fmOxide, level = 0:2)

you get the predicted outcomes for each level of grouping, i.e. fmOxide$fixed.