I'm analysing PISA-data, whitch uses Fay's replicated weights. I've managed to build a simple SEM and implementing REPSE=FAY (.5) with 80 replicated weights that are in the PISA data set, but I'm uncertain concerning witch estimator to use. Any advice? ML doesn't seem to give any warning message in the output, but many others do, adn refer to a unspecified "default estimator".
My second quiestion concerns the S.E. for the standardized model. These do not appear as they do for the unstandardized model, but I can't find any specific error/warning message concerning this.
My analysis syntax is as follows:
ANALYSIS: TYPE = COMPLEX; ESTIMATOR IS ML; ITERATIONS = 1000; CONVERGENCE = 0.00005; COVERAGE = 0.10; REPSE = FAY (.5);
The estimator you should use is ML. The estimator specification is for the point estimates only since the replicate weights determine the standard errors (thus the MLR/MLF estimators are identical to ML in this case - these estimators yield the same point estimates as ML in all situations).
The replicate weights standard errors for the standardized coefficients are not available yet. If you can write the standardized parameters expressions in model constraints you can get the standard errors.