10.07.2015 Views

Mplus Users Guide v6.. - Muthén & Muthén

Mplus Users Guide v6.. - Muthén & Muthén

Mplus Users Guide v6.. - Muthén & Muthén

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Examples: Multilevel Modeling With Complex Survey DataCHAPTER 9EXAMPLES: MULTILEVELMODELING WITH COMPLEXSURVEY DATAComplex survey data refers to data obtained by stratification, clustersampling and/or sampling with an unequal probability of selection.Complex survey data are also referred to as multilevel or hierarchicaldata. For an overview, see <strong>Muthén</strong> and Satorra (1995). There are twoapproaches to the analysis of complex survey data in <strong>Mplus</strong>.One approach is to compute standard errors and a chi-square test ofmodel fit taking into account stratification, non-independence ofobservations due to cluster sampling, and/or unequal probability ofselection. Subpopulation analysis is also available. With samplingweights, parameters are estimated by maximizing a weightedloglikelihood function. Standard error computations use a sandwichestimator. This approach can be obtained by specifyingTYPE=COMPLEX in the ANALYSIS command in conjunction with theSTRATIFICATION, CLUSTER, WEIGHT, and/or SUBPOPULATIONoptions of the VARIABLE command. Observed outcome variables canbe continuous, censored, binary, ordered categorical (ordinal), unorderedcategorical (nominal), counts, or combinations of these variable types.The implementation of these methods in <strong>Mplus</strong> is discussed inAsparouhov (2005, 2006) and Asparouhov and <strong>Muthén</strong> (2005, 2006a).A second approach is to specify a model for each level of the multileveldata thereby modeling the non-independence of observations due tocluster sampling. This is commonly referred to as multilevel modeling.The use of sampling weights in the estimation of parameters, standarderrors, and the chi-square test of model fit is allowed. Both individualleveland cluster-level weights can be used. With sampling weights,parameters are estimated by maximizing a weighted loglikelihoodfunction. Standard error computations use a sandwich estimator. Thisapproach can be obtained by specifying TYPE=TWOLEVEL in theANALYSIS command in conjunction with the CLUSTER, WEIGHT,WTSCALE, BWEIGHT, and/or BWTSCALE options of the233

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!