Measurement Bias Detection Through Factor Analysis
M. T. Barendse a , F. J. Oort a , C. S. Werner b , R. Ligtvoet a & K. Schermelleh-Engel
The study compares the performance of MGFA, RFA with LMS, and RFA with RSP in detecting uniform and nonuniform measurement bias, with respect to both dichotomous and continuous violators, in both single run and iterative procedures.
16 conditions
4 types of bias: no bias/uniform bias/nonuniform bias/uniform and nonuniform bias
2 types of violators: continuous/ dichotomous
2 relationships between trait and violator: independent or dependent
Mplus (version 5.2; Muthén & Muthén, 2007) is used to analyse the data.
The results show that RFA methods outperform MGFA when the violating variable is continuous.
Comments:
It is an interesting and useful article for empirical study.
It seems that if the violators are dichotomous such as different gender or groups, the results of the three methods are similar. And the results of iterative approaches for the three methods are of few differences.
What about the polytomous distractors?
What will happen if the items are not continuous response scales, should we choose IRT methods or just treat them as continuous?