Evaluating the Performance of Three Structural Equation Models With a Small Number of Indicators: A Monte Carlo Study

• In Event: 63.085 - Poster Session 13
  In Poster Session: 63.085-4 - Modeling Nonindependent Data

Abstract

This simulation study examines the performance of three SEM models when latent factors have a small number of indicators and sample size is small. The manipulated factors include sample size (of 50, 80, 100, 150, and 200), factor loading (.6 as low, .8 as high), and regression coefficients (γ and β) in the structural model (0.6 and 0.5 as high and 0.3 and 0.2 as low). The preliminary simulation showed that structural regression model has the best performance in term of accuracy while path analysis with factor scores shows the worst performance. This result also shows that all three simulated factors have the significant effect on performance of these methods.

Authors

• Diep Thi Nguyen, University of South Florida

• Eun Sook Kim, University of South Florida

• Thanh Vinh Pham, University of South Florida