Does anyone know a method how to calculate the required sample size in a repeated measures design that will be analysed using linear mixed modelling? The design looks as follows:I have two groups (intervention/control) that are assessed at four time points (t0 before the intervention, t1 after the intervention, t2 after a three-month maintenance phase and t3 at 6-month follow-up). In the linear mixed model I will look at the interaction of time and group, allowing for random slopes and intercepts.

I will make the following assumptions:

Effect size for the comparison of groups at t2, adjusted for baseline values: d=0.45

variability of means: sd= 0.75

power: 0.8

alpha level: 0.05

  • $\begingroup$ Before you can do such analysis you need to specify how big an effect you wish to detect and the variability of the measurements. This becomes more of a problem as you add interaction terms and repeated measures over time. See this page for available tools, this page for an alternate approach with simulation, and this page for the importance of some assumptions. You will need to provide more information about your study to get a useful answer. $\endgroup$
    – EdM
    Feb 28, 2019 at 14:46
  • $\begingroup$ check the powerlmm package in R: cran.r-project.org/package=powerlmm $\endgroup$ Feb 28, 2019 at 19:06


Your Answer

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