서강대학교

초록

Previous research shows mixed results on how ignoring measurement non-invariance in factor loadings affects the estimation of path coefficients. In attempts to reconcile these conflicting findings, I probed into whether the total sum of measurement non-invariance in factor loadings critically impacts the estimation of path coefficients for the groups of interest. I proposed that as the sum of non-invariance in factor loadings approaches zero, a bias in path coefficients would become attenuated. To test our proposition, I conducted a Monte Carlo study in which the overall sum of non-invariance was set to systematically vary. The results supported our proposition: bias in path coefficient was reduced as the sum of non-invariance approached zero and was in fact the smallest when the sum of non-invariance equaled to zero. These findings point to total sum of non-invariance as a potential mechanism that can account for previously mixed results related to the effects of ignoring measurement non-invariance on path coefficients. It also calls researchers to pay greater attention to the issue of measurement invariance itself, as path coefficients may appear unbiased even when measurement invariance has been violated.