Anonymous posted on Thursday, February 06, 2003 - 10:34 am
I am using data that were sampled in the following way: all zip codes within a city were delineated and then a sample proportional to size of the zip code was selected. I will be using EFA. Is there a way to check if there is clustering within the zip codes? For example is there a way to compare EFA run with and without the cluster option?
bmuthen posted on Thursday, February 06, 2003 - 10:50 am
Mplus does not do multilevel EFA, but there are ways around that. One approach is to use Mplus Type = Twolevel to get the pooled-within sample covariance matrix and use the corresponding correlation matrix as input for an EFA run. My experience is that this matrix gives estimates that are close to within-level estimates of multilevel analysis.