I am working on analyzing the data from a cluster randomized trials. A total of 14 schools participated -- 7 intervention and 7 control. The randomization was done in pairs of schools -- within each pair, one school was randomized to the intervention and the other to control. I was asked to run the analysis while preserving the "yolked" (i.e., randomization in pairs) structure of the design. How do I do this in Mplus?
I coded the 7 pairs and used the CLUSTER command, but I am not sure it is correct. Do I need to calculate ICCs? How do I do this? Also, how do I get the standard deviation for the constructs measured in each of the schools?
Randomized pairs sounds like matched pairs randomization for which there are special analysis techniques (preserving the yolked design). Such analysis is not done in Mplus. It is different from a completely randomized design which Mplus would assume.
Can Mplus be used to analyze cluster randomized controlled trial? I'm planning a cluster RCT study with 8 clinics and 30 participants from each clinic. 4 clinics will be assigned to intervention, and 4 clinics will be assigned to control.
I have seen study protocols using Mplus to estimate power for similar design, but unsure how to set this up in Mplus in terms of clustering and ICC.