Course syllabus

Lane Fischer Richard Sudweeks

340-L MCKB 150-M MCKB

801-422-8293 801-422-7078

Email: lane_fischer@byu.edu Email: richard_sudweeks@byu.edu

Office Hours: Office Hours:

????????? T, Th, 10 a.m.–-11. a.m.

Wed., 10 a.m.—3 p.m.

T & Th, 12:00—2:50 p.m.

102 SWKT
Required Textbooks

Geiser, C. (2010). Data analysis with Mplus. New York: Guilford.

Kline, R.B. (2011). Principles and practice of structural equation modeling (3^rd ed.). New York: Guilford.
EXPECTED LEARNING OUTCOMES

As a result of completing this course, students should be able to:

1. 
   Explain the similarities and differences among (a) path analysis, (b) confirmatory factor analysis, (c) analysis of structural models, and (d) longitudinal growth curve analysis.
   
2. 
   Use confirmatory factor analysis procedures to assess alternative measurement models describing hypothesized relationships between latent variables and their respective indicator variables.
   
3. 
   Use confirmatory factor analysis to investigate questions related to reliability of measurement, construct validity, and factor invariance either across multiple groups of persons or with a single group across multiple occasions.
   
4. 
   Test alternative structural models describing hypothesized patterns of relationships among endogenous and exogenous variables including direct effects as well as mediating and moderating effects.
   
5. 
   Demonstrate proficiency in using SEM software (e.g., AMOS, Mplus, & LISREL) and interpreting the resulting output.
   
6. 
   Prepare written reports of completed SEM studies in accordance with commonly accepted guidelines and recommendations.
   
7. 
   Critique published reports of completed research and evaluation studies that utilized a SEM approach.
   

SEM is a family of methods that collectively provide a comprehensive data analytic framework for testing models that describe hypothesized relationships among sets of observed and latent variables. SEM is an extension of multiple regression, path analysis, and exploratory factor analysis that subsumes these narrower approaches as well as analysis of variance into a broader, more general framework. SEM can be used to test hypotheses about differences in group means as well as hypothesized associations among pairs of variables, but in SEM the focus is on testing alternative models that represent theories describing hypothesized patterns of relationships among sets of variables.

SEM can be used to analyze survey data, archival data, and other forms of nonexperimental data as well as data from experimental designs. SEM improves upon path analysis by allowing for the inclusion of latent variables as well as observed variables. In addition, SEM improves upon multiple regression by providing explicit estimates of measurement error and producing parameter estimates that describe relationships among latent variables after they have been adjusted for the estimated measurement error.

The roots of SEM can be traced back to developments in path analysis during the 1920s and 1960s, factor analysis in the 1920s and multiple regression during the 1950s and 60s. But the unifying framework and thee software that made SEM feasible were developed by Karl Joreskog and his colleagues in the 1960s and 1970s. Since that time SEM has been used for a wide variety of purposes in the social and natural sciences. During the last 20 years its use has grown exponentially. One reason for this growth has been the development of more user-friendly software such as EQS, AMOS, and Mplus. Another reason has been the ongoing development of refinements in SEM concepts and procedures.

SEM includes two main types of models that are used sequentially. The first is the measurement model which has come to be known as Confirmatory Factor Analysis (CFA). It provides a way of testing whether a designated set of observed variables can be interpreted as one or more valid and reliable latent variables. The second type is the structural model that is used after a measurement model has been found to fit the data and which provides a way of testing hypothesized relationships among the latent variables defined by the measurement model.

SEM provides a language and a set of concepts for thinking about and designing research. It also provides a set of procedures and conventions for analyzing, interpreting, and reporting the results of research. Research journals in education as well as in psychology, sociology, and other related disciplines now regularly publish reports of studies based on SEM.