A comparison of sequential and nonsequential specification searches in testing factorial invariance

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Factorial invariance testing is crucial for cross-group comparisons.
• Modification indices (MIs) are used to identify model misspecifications.
• Two common approaches for using MIs are nonsequential and sequential search.

Purpose of the Study:

• To evaluate the performance of modification indices (MIs) in identifying correct partial invariance models.
• To compare the nonsequential and sequential search methods for testing factorial invariance.

Main Methods:

• Simulated data were used to test factorial invariance across groups.
• The nonsequential search involved relaxing all MIs above a cutoff simultaneously.
• The sequential search involved relaxing one parameter with the largest MI at a time.

Main Results:

• The nonsequential search method resulted in high false positive rates, incorrectly identifying invariant items as noninvariant.
• The sequential search method demonstrated good true positive and false positive rates.
• The sequential search method is recommended for accurate scale modification.

Conclusions:

• The sequential search method is superior to the nonsequential method for testing factorial invariance using MIs.
• Accurate identification of partial invariance is essential for valid scale modification and cross-group analysis.