EFA with strata PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 Ann-Renee Blais posted on Friday, February 27, 2015 - 7:27 am

I have weighted survey data (6 strata) on which I want to run an EFA. In order to account for the weights, I know I simply add "WEIGHT IS (weight variable)" However, how do I take into account the stratified nature of the data, especially with only 6 strata?

Would a MIMIC model with 5 dummy-coded variables be the best way to proceed? Or could I get away with "TYPE IS COMPLEX" and "STRATIFICATION IS (stratum id)"?

Best regards,

Ann-Renee Blais
 Tihomir Asparouhov posted on Friday, February 27, 2015 - 3:36 pm
You will NOT have any problems because of small number of strata.
 Ann-Renee Blais posted on Friday, February 27, 2015 - 3:53 pm
Thank you, Tihomir, for your prompt response!

I was under the impression too few strata would be as problematic as too few clusters? Would TYPE IS COMPLEX still be appropriate with 6 clusters? How are the 2 situations different?
 Tihomir Asparouhov posted on Friday, February 27, 2015 - 7:30 pm
With 6 clusters it would be a problem but not will 6 strata.

The two situations are different in the following way. Stratified sampling reduces standard errors while cluster sampling increases standard errors.

If you drop the strata variable from the analysis - the analysis is as if there is only one stratum (the whole population is one stratum). So adding the strata variable to the analysis will reduce further the SE and should not be causing problems the way small number of clusters would.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message