ABSTRACT
Let’s review our progress in our adventures beyond MR. You know how to conduct path analysis using MR. This experience includes the estimation of standardized and unstandard-
ized paths, the calculation of disturbances ( 1 2−R ), and the calculation and comparison of direct, indirect, and total effects using two different methods. We transitioned into estimating path models using Amos and other SEM programs and focused again on the estimation of both standardized and unstandardized effects and direct, indirect, and total effects. With Amos, we switched from the estimation of the paths from disturbances to estimating the variances of the disturbances, although either is possible. We have defined just-identified, overidentified, and underidentified models, and I suggested that you use a SEM program to estimate overidentified models but use either MR or an SEM program if your models are just-identified. We have examined fit indexes for overidentified models and have highlighted a few that are useful for evaluating a single model and those that are useful for comparing competing models. We briefly focused on equivalent models, nonrecursive models, and longitudinal data. We focused on the effects of measurement error on path analysis, MR, nonexperimental research, and research in general and began considering the use of latent variables as a method of obviating this threat. We expanded our knowledge of latent variables, their meaning, and estimation via confirmatory factor analysis.
