I've been trying to find out if it would be possible to use linear mixed models on repeated measurements data when there are only two measurements within each participant. This paper (doi: 10.1002/sim.2538) and related papers seem to argue that this is not possible. I can't really figure out why? Most posts I find on cross-validated don't seem to explicitly mention that this would be an issue... Any help?

  • $\begingroup$ As a side note it is very feasible to use generalized least squares or a first-order Markov model on only 2 measurements per participant. $\endgroup$ Feb 21 at 13:18
  • $\begingroup$ You could always run a simulation and power analysis to determine how effective such a model can be to detect the effects you are interested in. Whether or not there will be issues depends on the data (eg how well does it fit the model? How large is the variability?) $\endgroup$
    – bdeonovic
    Feb 21 at 15:49


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy