Model fitting procedures for nonlinear factor analysis using the errors-in-variables parameterization
Traditional factor analysis and structural equation modeling use models which are linear in latent variables. Here, a general parametric nonlinear factor analysis model is introduced. The identification problem for the model is discussed, and the errors-invariables parametrization is proposed as a solution. Two general procedures for fitting the model are described. Tests for the goodness of fit of the model are also proposed. The usefulness and comparison of the model fitting procedures are studied based on a simulation.
KeywordsNonlinear Factor Error Variance Estimate Factor Vector Approximate Likelihood Maximum Normal Likelihood
Unable to display preview. Download preview PDF.
- Amemiya, Y. (1993a). Instrumental variable estimation for nonlinear factor analysis. In Multivariate Analysis: Future Directions 2. (C.M. Cuadras, C.R. Rao, Ed.), pp. 113–129, North-Holland Series in Statistics and Probability, Amsterdam.Google Scholar
- Amemiya, Y. (1993b). On nonlinear factor analysis.Proceedings of ASA Social Statistics Section290–294.Google Scholar
- Jöreskog, K. G. (1977). Factor analysis by least squares and maximum likelihood methods. InStatistical Methods for Digital ComputersVol. 13, K. Enslein, R. Ralston, and S.W. Wilf (eds.). Wiley, New York.Google Scholar
- Mooijaart, A. and Bender, P. (1986). Random polynomial factor analysis. InData AnalysisandInformatics IV(E. Diday et al., Ed.), pp. 241–250. Elsevier Science Publishers B.V., New York.Google Scholar
- Yalcin, I. (1995). Nonlinear factor analysis. Ph.D. dissertation. Iowa State University.IowaGoogle Scholar
- Yalcin, I. and Amemiya, Y. (1993). Fitting of a general nonlinear factor analysis model.Proceedings of ASA Statistical Computing Section92–96.Google Scholar