A solution to the problem of measurement error is introduced. It is agreed that most research is oversimplistic in assuming perfect measurements. The solution that consists in using partial correlations is shown to be only occasionally successful. Typically, multiple indicators are needed, & once there are at least as many known as unknown r coefficients, all path coefficients can be estimated. Tests of the global theory can be carried out, once there are more knowns than unknowns. It is pointed out that, although the procedure has been explained in terms of r coefficients & path coefficients, this is unlikely to be the best procedure when making comparisons between populations or when overidentified systems are dealt with. 5 Model Diagrams. A. Orianne.
