ABSTRACT
Herbert L. Costner, H. M. Blalock, and D. Heise have demonstrated the application of path analysis to problems involving multiple indicators of underlying constructs. This chapter discusses the use of a general method for factor analysis (Karl G. Joreskog) to obtain a single "best fit" estimate of each parameter and an overall goodness of fit test for the consistency of the data with the model. Joreskog's confirmatory factor analysis procedures allow the investigator to fix some parameters of the model and estimate the others. To formulate the model in terms of Joreskog's general model, one has to write a structural equation for each observed variable. The Estimating Models Involving Unmeasured Variables variance-covariance matrix of the factors is assumed to be the population variance-covariance matrix. In practice, the population matrix is not available and the parameters have to be estimated from a sample variance - covariance matrix.
