[R-sig-ME] Random Slope for Dummy Variables

Junyan Luo jzl106 at gmail.com
Thu Apr 7 23:42:33 CEST 2011


Hi Doug,
Thanks for your reply. I think the "natural formulation" of factor
variables in R is essentially treating one level as reference while
regarding other levels as dummies, right? So here the use of dummy
variable is really an attempt to simplify the question (i.e. restrict
the factor to 2 levels).
That being said, the same problem of uneven grouping is still there. I
know I can run the model easily in R, but my question is whether
that's statistically valid. Even if we use your 5-level categorization
of ethnicity, what if those ethnic groups are distributed extremely
unevenly among different schools? What if one ethnic group are not
present in some schools at all? What the random slope means in those
cases?
Again, thank you for your attention!
REGARDS,
J. Lo




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