Using SEM to Analyze Complex Survey Data: A Comparison between Design-Based Single-Level and Model-Based Multilevel Approaches

The effect of individual differences in the inner and outer states of ICT on engagement in online reading activities and PISA 2009 reading literacy: Exploring the relationship between the old and new reading literacy

“…The two-stage sampling scheme of PISA data should be carefully dealt with to have consistent and correct statistical inference of parameter estimates (Wu and Kwok, 2012). We performed the mediation analysis (MacKinnon, 2008) with replicate weights and plausible values to prevent biased population parameter estimates, to preserve the nature of the PISA two-stage stratified sampling scheme, and to approximate each individual's true score/ability (OECD, 2009b) using Mplus 6.11 (Muthén & Muthén, 2010).…”

Reflections on Partial Least Squares Path Modeling

“…Chin & Dibbern, 2010;Crisci & D'Ambra, 2012) would also need to be adjusted for complex survey data (Pesarin & Salmaso, 2010). Guidance for incorporating MLM and design-based resampling techniques into PLS-PM can be found in the SEM field, which has made considerable advances in both of these areas (e.g., Bai & Poon, 2010;Hox, 2013;Kaplan, Kim, & Kim, 2009;Oberski, 2013;Rabe-Hesketh, Skrondal, & Zheng, 2007;Stapleton, 2008;Wu & Kwok, 2012).…”

Why and when workplace ostracism inhibits organizational citizenship behaviors: An organizational identification perspective.

“…To deal with the non-independence data due to the nested structure of performance ratings (i.e., 150 supervisors rated 282 subordinates), we used a design-based modeling approach that "takes the multilevel data or dependency into account by adjusting for parameter estimate standard errors based on the sampling design" (Wu & Kwok, 2012, p.17) (TYPE = COMPLEX, ESTIMATOR = MLR in Mplus). This design-based modeling approach is appropriate for our research because it handles non-independence data structures when mechanisms at a single level (i.e., employee level in this study) are examined (Wu & Kwok, 2012). We estimated two path models with the composite scores of our research variables.…”