Abstract

The structural equation modeling (SEM) framework is an increasingly popular data analytic method in the social sciences (MacCallum and Austin 2000; Martens 2005). The inherent flexibility of the SEM framework and developments in theory (e.g., Bollen et al. 2010;Muthén 2002) and accompanying software (e.g., lavaan, Rosseel 2012; LISREL, Jöreskog and Sörbom 1996–2001; Mplus et al. 1998–2010) allow analysts to employ complex methods and have contributed to growth of use. Of particular interest for this chapter are advances that accommodate the peculiarities of survey data in general and educational achievement data in particular. To that end, the purpose of this chapter is to demonstrate the use of SEM in analyzing international large-scale assessment data. We address issues surrounding the treatment of missing data, deciding whether and when sampling weights should be used, estimator choices, and handling plausible values. The subjects of missing data, sampling weights, and plausible values are covered in greater detail in other chapters in this volume (see Chapters 6, 8, 17, and 20 of this volume). Thus, we avoid an in-depth overview of these matters and instead concentrate on employing methods that deal with these issues in the SEM framework. We also briefly overview estimators that are well suited for SEMs and complex data. Our examples include both single- and multilevel models.