Comparing Multiple Indicators Multiple Causes (MIMIC) and Alignment Method for Investigating Measurement Invariance With Respect to a Continuous Variable

• In Event: Advancements in Detecting Differential Item Functioning and Measurement Invariance

Abstract

In social sciences research it is very common, and important to have comparisons over continuous variable (e.g., depression across years of education). Arbitrary categorizing was used to create nominal variable for multiple group comparison, which could compromise statistical power(Young, 2016). In this paper, we compare the performance of multiple indicators multiple causes (MIMIC) and the alignment method in measurement noninvariance detection, when a violator is a continuous variable. We conducted a Monte Carlo simulation under various conditions. Preliminary result showed that both MIMIC and alignment method have good power and lower type I error rate when the violator is linearly related with noninvariant item; MIMIC is not sensitive to nonlinear violation of noninvariance.

Authors

• Yuanfang Liu, University of Cincinnati

• Hok Chio (Mark) Lai, University of Southern California

• Benjamin Kelcey, University of Cincinnati