Phillip Gee posted on Tuesday, March 22, 2011 - 5:07 am
I've recently started using MPlus to conduct some multi-level analyses. My experiment involves 30 participants in a repeated measures design. Each participant performs on a task over 10 trials, and they are measured on several variables between trials. Each trial lasts for 5 minutes, and participants are also measured on several other variables at five intervals within trial.
The design is 30 between-person independent x 10 between-trial repeated measures x 5 within-trial repeated measures. Each participant provides a total of 10 responses on the between-trial measures, 50 responses for the within-trial measures, and once on between-person measures. We expect responses at both the between- and within- trial levels to vary systematically over time.
I am interested in the within-trial level of data, and I am looking to conduct multi-level SEM/path analyses. I understand that I need to account for the higher levels of variability. If I want to test relationships among within-trial variables, I need to account for the variability of these variables between trials (growth/clustering between trials) and between people (clustering between people) simultaneously. Ideally, I would like to setup a within-trial level model that can test for cross-level interactions at between-trial and between-person.
I'd really appreciate any advice on how to analyse this properly in MPlus!
Perhaps you can do it as a type=twolevel analysis in the following way.
Let the within-trial variables form multivariate response (so columns in your data). The model captures the correlations among the variables across the 5 within-trial repeated measures. Repeat these data on different records for the different trials. Repeat over the 30 subjects.
In this way, level 1 is variation over trials and level 2 is variation over subjects.