[R-sig-ME] Model specification help

Andrew Perrin clists at perrin.socsci.unc.edu
Thu Mar 8 21:50:00 CET 2007


I am trying to estimate a large mixed-effects model. The data consist of 
all grades issued to all undergraduates at UNC in the last ten years -- as 
you can imagine, a fairly large set! What I want to estimate is the 
relative effects of student performance, instructor practices, and 
departmental practices.  Here's what I tried:

grades.lmer<-lmer(formula   = grade.pt ~ cour.dep + section +
                   (stud.id + instr.id | section/instr.id/cour.dep),
                   na.action = na.omit,
                   data = newgrades.stripped.df)

I get the following:

Error in array(0, c(n, n), list(levs, levs)) :
         'dim' specifies too large an array

My *intention* in this is that grade.pt (the numerical grade issued) is 
modeled as a function of course.dep (the department, a fixed effect); 
section (the specific course section taken, a fixed effect); and random 
effects for stud.id (individual student) and instr.id (individual 
instructor), where grades are nested within sections, nested in turn 
within instructors, nested in turn within departments.

I would welcome both substantive and technical advice here. If the problem 
is simply that the dataset is too big, I'd be OK with taking a random 
sample of it; but if there's something else wrong I'd be grateful for your 

Andy Perrin

Andrew J Perrin - andrew_perrin (at) unc.edu - http://perrin.socsci.unc.edu
Assistant Professor of Sociology; Book Review Editor, _Social Forces_
University of North Carolina - CB#3210, Chapel Hill, NC 27599-3210 USA
New Book: http://www.press.uchicago.edu/cgi-bin/hfs.cgi/00/178592.ctl

More information about the R-sig-mixed-models mailing list