Which model to use ?


I have a data coming from the ophthalmology field, i.e, eyes.

For each patient, an intervention is made on one eye (either left or right, it is chosen randomly), and some measure is taken and compared with the measure of the second eye which is the control. Of course, these measures are also taken prior the intervention. In addition, these measures are being checked in a few more time points, meaning that this is a longitudinal data.

What bothers me is, that unlike a "simple" longitudinal data, in which you have a treatment group and control group being screened over time, here the treatment and control are paired, from the same person.

I know how to do paired analysis, I know how to approach longitudinal, but how do I combine them ? My measure is continuous, and I have a few more background covariates which are not important right now.

Thanks !


Fortran must die
If you compared the change in the difference of the two eyes over time than I would think the differences would be paired - and you could do repeated measures. But why do controls this way? I think more traditional would be to have whatever you measure, measured separately for the control group and for the intervention group and see if the intervention changed with repeated measures (and the control group did not of course).
Let me see that I understand you. You say that if I work with D=X-Y, then I already handle the pairing problem, and I can treat my data as I would for any longitudinal study ?

You are correct about having a "normal", treatment-control design, however this is a preliminary study where the control is the second eye, a kind of a feasibility study. And the sponsor is saving money... :)


Fortran must die
That is what I was suggesting. But I would look for a second opinion, or in the literature for this, since I am anything but an expert. Most of what I do is in regression and time series.