This paper addresses the problem of error rate inflation when conducting a sequence of model modifications. The problem is identified as increasing the probability of a Type II error. It is argued that the most effective way of controlling the inflation of the Type II error in the covariance structure modeling context is to choose modifications that have maximum power. Further, it is argued that methodology based on expected parameter change allows researchers to modify models in accordance with substantive considerations as well as minimize the inflation of the Type II error rate.
Get full access to this article
View all access options for this article.
References
1.
Buse, A. (1982). The likelihood ratio, Wald, and Lagrange multiplier tests: An expository note. The American Statistician, 36, 153-157.
2.
Cliff, N. (1983). Some cautions concerning the application of causal modeling methods. Multivariate Behavioral Research, 18, 115-126.
3.
Jöreskog, K. G. and Sorbom, D. (1984). LISREL-VI: Analysis of linear structural relationships by the method of maximum like-lihood. Mooresville: Scientific Software, Inc.
4.
Kaplan, D. (1988, April). Modification of structural equation models: Application of the expected parameter change statistic. Presented at the meeting of the American Educational Research Association, New Orleans, LA.
5.
Kirk, R. E. (1982). Experimental design: Procedures for the behavioral sciences (2nd ed.). Belmont: Brooks/Cole.
6.
Luijben, T. , Boomsma, A., and Molenaar, I. W. (1987). Modification of factor analysis models in covariance structure analysis: A Monte Carlo study. Heymans Bulletins Psychologische Instituten, R. U. Groningen.
7.
Saris, W. E. , Satorra, A., and Sorbom, D. (1987). The detection and correction of specification errors in structural equation models. In C. C. Clogg (Ed.), Sociological Methodology, 1987. San Francisco: Jossey-Bass.
8.
Satorra, A. (in press). Alternative test criteria in covariance structure analysis: A unified approach. Psychometrika.
