[R] mixed model question and using lmer

Wiggin wiggin.peters at gmail.com
Thu Jun 2 02:24:29 CEST 2011


i am analyzing some data and have a question i hope someone can
answer.

i want to use this sort of model:

lmer( y ~ x + (1 | ID ), family=binomial, weight=w)

so i want to explore the relationship between y and x, with a random
effect for each patient.

my question is this.  is this a sensible model to use when the number
of observations for each patient ID is often one? I have 305
observations for 239 patients.186 of the patients have only one
observation each and another 40 have two observations each.   at the
other extreme, one patient has 5 observations

lmer fits the model without complaining.

but is this a sensible thing to do?  it might be silly if so many of
my patients only have on observation each.  i've tried doing some
reading but i cannot find an answer to this question.

thanks you for your advice

wiggin.



More information about the R-help mailing list