Dear all,
I have repeated measures over time on a number of unique individuals where
each observation was matched to 5 controls (i.e., 1:5 matched case
control). If the repeated measures are independent I intend to use
conditional logistic regression to contrast each observation with the
corresponding controls (i.e., cluster on each unique observation).
However, if the repeated measures are correlated I believe I should adopt
a random effects model (either subject specific or population averaged)
and cluster on the individual. A random effects model would force me to
forfeit my matched case-control design.
I have been to the texts and primary lit. and the majority of discussion
around repeated measures specific to generalised linear models refers to
cohort studies. I would be eternally grateful if someone provided
guidance/suggestions/further readings for a logit model that accommodates
repeated measures and matched case-control data.
Thanks.
Chris
|