Hi, i want to ask two question about cluster analysis. In MASEM analysis, i made a cluster analysis according to sample type (older, younger). And the homogenity test which assumes correlation matrix between the studies was rejected.
i) i continued the second step of MASEM. And model fit indices were good. Does it make sense to continue analyzes and test models even though the correlation matrices are heterogeneous? Are the model fit indices and parameters I obtained reliable?
ii) The Q statistic shows that subgroups do not adequately explain the heterogeneity. So what percent of the variance between studies is explained by this moderator variable? Does MASEM give a statistic (Qbetwwen, Qwithin etc.) for it?
Printer-friendly version
Hi vildan,
Q1. This is a standard multiple-group SEM on the correlation matrices. I think that you can rely on what we have learned in interpreting SEM models.
Q2. It is not easy to define a percentage of variance explained in a complicated model like SEM. If you want, you can define a Q statistic on the between and within effects by comparing models with and without the constraints. But you may have to work it out as the metaSEM package does not report it.
Mike