We are conducting multi-level modeling with Mplus 5.2 with repeated measures data (cognitive processes in 3 different situations, i.e., situations nested within individuals). However, we also measured a friendís comment within each situation (i.e., 3 types of comments within each of the 3 situation types). So far, we have been estimating 2 separate 2-level models for this data (i.e., either situation or comment nested within the individuals). There are 2 things that we are not sure of. 1) There seems to be an issue with the situation dummy variables at the within level. We are getting errors referring to the singularity of the covariance matrix for the third dummy variable and the model doesnít terminate correctly. We have been trying to modify different statements in the syntax (e.g., estimator/variable definitions). Can you tell us what specifically needs to be taken into account in the syntax when including dummy variables at the within level? We also use FIML estimation of missing data. 2) Would it be statistically correct to construct a 3 level model where the 3 different friendís comments are nested within the 3 different situations (constructing dummy variables such that each participant is represented in the data 9 times) which in turn are nested within individuals? We would appreciate any references which might illustrate the appropriate approach/model/syntax for this type of RM data structure.
In Mplus, it is not necessary to use TYPE=TWOLEVEL; for a growth model. Mplus takes the wide multivariate approach to growth modeling. See examples in Chapter 6 of the user's guide. Please see also the introduction to Chapter 9. If you get error messages in the output, please send the output and your license number to firstname.lastname@example.org.