Hello, I have a question about the scaling of factors in a CFA/SEM with clustered data. As far as I understand, it is probably not a good idea to fix the factor variance to 1.0 when using the type=twolevel - option, since the variance could differ between the clusters. But what about the type=complex - option? Would it be reasonable to fix the variance instead of fixing a factor-loading in this case (since there is no disaggregation)?
bmuthen posted on Monday, October 31, 2005 - 12:37 am