Characteristics of Structural Equation Models which Affect the Power of the Likelihood Ratio Test

In the early days of causal modelling in the social sciences a lot of attention has been paid to testing of the models (Simon, 1954; Blalock, 1962; and Goldberg, 1966). Due to the fact that these procedures only provided tests of the models but no estimates of the causal effects, the attention shifted in the next decade to the estimation of the causal effects (Boudon, 1965; Duncan, 1966; Land, 1969). In this approach OLS-estimation is used. But using this procedure the testing of the models requires a lot of extra work which is in general ignored. As a consequence now many causal models are still published in the social science journals which do not fit the data analysed (Saris et al., 1985).