Factor analysis on complex survey data

[byrne48]

Can I preform factor analysis for complex survey data? (I need to include a weight, cluster, strata, I also need a domain statement)

Any help with code for this would be much appreciated 

Thanks!

[PaigeMiller]

there is no SURVEY procedure for factor analysis.

I think you would have to compute the proper weights for each observation and then run PROC FACTOR with the WEIGHT statement.

[byrne48]

Thanks for your response!

Just to clarify on computing proper weights  - each of my participants have an individual weight variable, is this what I would use for my weight statement? 

Can a domain statement be used in proc factor? 

Thanks!

[PaigeMiller]

Each record in your data step has to have a weight. You would have to compute the weight for each record based upon the design of the survey. There is no DOMAIN statement in PROC FACTOR and so if you want that functionality, you would have to program that yourself.

[djmangen]

I suspect that the OP is working with a complex survey design and is looking for the equivalent of Proc SurveyReg.  If this is the case, at present there is NOT a Proc SurveyFactor.

Some have examined a backdoor sort of approach -- going first to correlations.

https://jesshampton.com/white-paperspresentations/proc-surveycorr/

You could then create the correlation/covariance matrix that you want and feed that into Proc Factor.  I have contemplated doing this, but have never done it.

David