Improving the Evaluation of Construct Change Over Time: Advantages of Longitudinal Moderated Nonlinear Factor...

Area of Science:

• Psychometrics
• Longitudinal Data Analysis
• Developmental Psychology

Background:

• Conventional growth curve models assume stable measurement over time, which is often violated.
• Ignoring differential item functioning (DIF) can lead to biased estimates of construct change.
• Existing methods often rely on restrictive assumptions like item equality.

Purpose of the Study:

• To introduce and validate a novel second-order growth model that accounts for time-varying and time-invariant DIF.
• To assess the model's performance in detecting DIF across different groups and over time using simulations.
• To demonstrate the model's utility in analyzing real-world developmental data.

Main Methods:

• Proposed a longitudinal extension of moderated nonlinear factor analysis (MNLFA).
• Incorporated Bayesian regularization for simultaneous evaluation of DIF from multiple covariates.
• Conducted a simulation study to test the model's DIF detection capabilities and bias reduction.

Main Results:

• The proposed MNLFA-based growth model effectively detects DIF without requiring anchor items.
• The new model avoids the biased growth estimates observed in conventional models.
• Simulation results confirm the model's ability to handle DIF in longitudinal data.

Conclusions:

• The flexible second-order growth model provides a more accurate approach to modeling construct change when DIF is present.
• This method enhances the reliability of longitudinal studies by addressing measurement invariance.
• The approach is applicable to various fields, including the study of child externalizing behaviors.