A less known methodology for measurement equivalence studies is Multiple-Indicator, Multiple Causes (MIMIC) analysis, a structural equation modeling (SEM) approach. This methodology does not require the large sample, and subsample, sizes needed for MG-CFA and IRT methods, can be used to test for DIF and DTF across more than two populations (or subpopulations) simultaneously, to test for DIF or DTF with continuous variables, and can be used to simultaneously control for multiple variables. This workshop will provide participants with the knowledge and practical skills needed to conduct MIMIC analyses to test for DIF and DTF using SPSS AMOS.
Methods: First the basic SEM MIMIC model for testing for DIF or DTF, and the theory underlying this model, will be presented and explained. Then participants will be walked step-by-step through a MIMIC model analysis testing for DIF in items on a Likert-type measure of suicidal thinking, and testing for DTF in the total scores on a standardized measure of self-esteem. This in-depth walk-through will show how to test for DIF and DTF across multiple populations simultaneously, and as a function of continuous scores such as age and those on a measure of depression. Participants will also be shown how to write up results of these analyses and how to do follow-up bias studies.
Results: Participants will know how to conduct and interpret the results of MIMIC analyses for DIF and DTF. Participants will also know how to do follow-up studies subsequent to the identification of DIF and/or DTF to determine whether bias exists at item score or test score levels; to identify sources of bias; to determine what can be done to eliminate bias; and how to write up the results of MIMIC analyses.
Conclusions and implications: Participants can expect to leave this workshop with a basic understanding of fundamental theory underlying the MIMIC models of DIF and DTF; skills in the use of AMOS for conducting MIMIC analyses; how to write-up results of MIMIC analyses; and the ability to use results from MIMIC analyses for developing measures that are unbiased and culturally and population sensitive.