Detecting item bias in latent construct between groop comparisons: an illustrative example using multi-sample covariance structural equations modeling