I came across a discussion thread on ResearchGate's where one researcher mentioned that repeated measure ANOVA is an outdated approach to longitudinal and intensive longitudinal data. It is known that drawbacks include order of treatments, often violated assumption of sphericity and according Kim of Analysis factor,

"One of the biggest problems with traditional repeated measures ANOVA is missing data on the response variable... problem is that repeated measures ANOVA treats each measurement as a separate variable. " Analysis Factor Blog

Is there a comparable or alternative best approach? (And please, I really need your insight and experience on this subject matter). Give reference, where your approach was explored with examples. Thanks.

  • $\begingroup$ See fharrell.com/post/re . Repeated measures ANOVA is indeed obsolete. $\endgroup$ May 5, 2022 at 12:37
  • $\begingroup$ I agree with you, @FrankHarrell. Apart from mixed effect models what other approaches are currently being used? $\endgroup$ May 5, 2022 at 13:05
  • $\begingroup$ See the above link. Primarily Markov models (which apply extremely widely) and generalized least squares (for continuous Y, Gaussian residuals). GEE is also an option, with cluster sandwich covariance correction, but is not as robust. $\endgroup$ May 5, 2022 at 13:19
  • $\begingroup$ @FrankHarrell, Oh, that is great! GEE , with advantage of consistent parameters estimates, is sure another option. Developed by (Liang and Zeger, 1986) to [produce regression estimates when analyzing repeated measures with non-normal response variables](publichealth.columbia.edu/research/population-health-methods/… ... ) $\endgroup$ May 5, 2022 at 14:03
  • $\begingroup$ GEE is a large sample procedure. Most of us don't have large enough samples where asymptotics can be trusted. $\endgroup$ May 5, 2022 at 18:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.