In the first class, students will be introduced to the most fundamental concepts in SEM. Additionally, students will be exposed to basic syntax required to run data in the Mplus 7 software package. 

Students will learn the principles of path analysis and the 5-step procedure for evaluating model-to-data fit. In doing so, students will become acquainted with the current conventions for reporting structural equation modeling methods within scientific manuscripts.

Students will be re-introduced to  factor analysis through the SEM framework using a technique referred to as “exploratory structural equation modeling” or “ESEM.” Next, students will be shown how to interpret and report ESEM output. Students will then have an opportunity to test CFA and ESEM syntax with their own data and sample datasets.  

 This class will emphasize traditional CFA and structural regression. We will take a closer look at some of the common problems users encounter when conducting measurement modeling and structural modeling. In addition, we will explore equivalent models, nested models, method effects, and MIMIC models. 

An important and often overlooked topic in SEM is what to do with missing data. This lecture will cover missing data assumptions, conventional and contemporary methods for handling missingness.

SEM requires a vast knowledge-base. There are tons of places where practitioners can make mistakes. One may make errors in syntax or one could encounter errors in model testing that point to problems in data entry, coding, etc. Most frequently, researchers inaccurately interpret SEM output including model fit indices, parameter estimates, and modification indices. This lecture will cover the most common mistakes and problems researchers will encounter and how to avoid them. 

To determine if measures are being interpreted the same way across groups (contexts or methods of delivery), researchers must learn the procedure for assessing group invariance. This procedure imposes constraints across groups in a sequential and progressive manner to test assumptions about equality of factor makeup, loadings, scales, and error. 

Students will learn how to apply invariance testing to measures assessed over time 

ENTER 

As we expand our knowledge of SEM and the endless possibilities, we begin to develop more complex questions about heterogeneity between and among cases within our datasets. Mixture modeling can be used to determine the existence of “latent classes” (i.e., subgroups) of individuals who may differ in levels across multiple outcomes (latent profiles analysis), or how they interpret items (factor mixture analysis). Individuals within a homogeneous group may not all start on the same page and/or they may differ in their rates of change, and this can be tested using growth mixture modeling 

SEM is generally considered a large sample technique. However, latent variable models can be tested with very small sample sizes (e.g., <50) within a partial least squares (PLS) framework. This lecture will cover the fundamentals of PLS modeling, parameter estimation and evaluation, and software implementation with SmartPLS