Testing for Measurement Invariance with many groups

Course Overview: We have witnessed a surge of cross-national surveys over the past few years. Large international surveys, like the European Social Survey or the World Values Survey, provide researchers with unique opportunities to test their theories and hypothesis in diverse populations around the world. However, this availability of data is only very seldomly accompanied by the realization that the assumption of comparability of the survey instruments should not be given but tested instead. Before attributing any relevant differences between populations to substantial theoretical reasons, methodological and measurement causes should be explicitly ruled out by testing for measurement invariance. This workshop will introduce participants to the basics of measurement invariance testing with many groups. We will start by explaining what is measurement invariance and the major causes for measurement non-equivalence in surveys. Then we will proceed to discuss the three most common approaches to measurement invariance testing and end with a simple tutorial on how to test for measurement invariance with Multi-Group Confirmatory Factor Analysis (MG-CFA) using R statistical software.

Content:

• Introduction
• The basic principles of measurement invariance testing
• The main causes of non-invariance
• The importance of measurement invariance testing in cross-national surveys
• Three most common approaches to measurement invariance testing
• Tutorial on MG-CFA two groups measurement invariance testing procedure in R
• Q & A

Objectives: By the end of the workshop, participants should be able to: a) Understand what measurement invariance is and why it is important in cross- national surveys. b) Recognize different alternatives to test for measurement invariance with many groups and their main strong points and limitations. c) Conduct a simple two-group measurement invariance testing procedure by running a Multi-Group Confirmatory Factor Analysis (MG-CFA) model using Lavaan (Rossel, 2012) and semTools (semTools Contributors, 2016) packages for R (R Core Team, 2018) statistical software.