Using Parcels to Convert Path Analysis Models Into Latent Variable Models.

Area of Science:

• Psychometrics
• Statistical Modeling
• Quantitative Psychology

Background:

• Measurement error can bias results in path analysis.
• Latent variable models offer a potential solution to this bias.
• Parcels can serve as indicators for latent variables in practical applications.

Purpose of the Study:

• To compare the efficacy of latent variable models using parcels versus path analysis models.
• To evaluate the impact of correcting path analysis models with reliability estimates.
• To determine the best approach for mitigating measurement error in structural equation modeling.

Main Methods:

• Latent variable models with parcels as indicators.
• Path analysis models using aggregated variables.
• Path analysis models corrected for measurement error using reliability estimates.

Main Results:

• Path analysis models yielded the smallest path coefficient estimates.
• Latent variable models with parcels produced the largest path coefficient estimates.
• Latent variable models demonstrated superior handling of measurement error.

Conclusions:

• Latent variable models utilizing parcels are preferable to path analysis with total scale scores.
• Employing parcels in latent variable models enhances the accuracy of path coefficient estimation.
• Researchers should prioritize latent variable modeling to overcome measurement error biases.