I am currently working on a paper with three levels. I have days (5) nested within persons nested within leaders. I want to test a mediation hypothesis about the influence of daily leadership behavior on daily employee job resources and daily employee work engagement. I am thinking about ways to analyze this. I am used to working with MLwin, but I heard something about growth curves or multilevel SEM with Mplus that could be useful for my design. My question is how you would recommend me to analyze my data?
Unless one of the levels is time, Mplus can estimate only two-level models. You could use TYPE=COMPLEX TWOLEVEL where for the third level standard errors are computed taking complex survey design features into account.