R help - Jun 2012 - lmer function in R

Dear R users,

I'm estimating a two-level regresion model but having difficuly in finding
a good R syntax example.

Let me use an example to explain what I'm doing. The dependent variable is
math score (mathscore). Predictors are at two levels. At the student level,
they are household income (hinc) and IQ (IQ). At the school level, there is
a dummy variable indicating if the school is a private school (prvsch). I
also have school IDs (id) that can be used as the subject.

I'm wondering what the syntax would be in R. Any advice is appreciated.

Best
Gary

	[[alternative HTML version deleted]]

Gary Dong <pdxgary163 <at> gmail.com> writes:
> I'm estimating a two-level regresion model but having difficuly in
finding
> a good R syntax example.
> 
> Let me use an example to explain what I'm doing. The dependent variable
is
> math score (mathscore). Predictors are at two levels. At the student level,
> they are household income (hinc) and IQ (IQ). At the school level, there is
> a dummy variable indicating if the school is a private school (prvsch). I
> also have school IDs (id) that can be used as the subject.
> 
> I'm wondering what the syntax would be in R. Any advice is appreciated.
  When using lme4 you don't need to explicitly state which fixed-effect
predictors are at which level; the simplest model would use

  mathscore ~ hinc + IQ + prvsch + (1|id)

You might want to consider whether the effects of household income
and IQ vary by school or not:

  mathscore ~ hinc + IQ + prvsch + (hinc+IQ|id)

(the intercept is implicit in all cases, this would be the same
as using (1+hinc+IQ|id))

  The full 3x3 covariance matrix is estimated in this case; if
you want to force independent variation of intercept, hinc, IQ
within schools, use 

  mathscore ~ hinc + IQ + prvsch + (1|id) + (0+hinc|id) + (0+IQ|id)

  Further questions on mixed models in R should probably go to
r-sig-mixed-models at r-project.org