Model Construction and Evaluation When Theoretical Knowledge Is Scarce

Publisher Summary During the past decade, many econometricians have come to realize that the use of much of the contemporary set of rigorous tools of statistical analysis is not applicable in all circumstances; this is mainly because the knowledge requirements of the investigator are far greater than are available. For example, a typical regression problem will require the researcher to know the relevant variables in the regression, its functional form, whether the observed relationship can be identified with the theoretical concept, and knowledge of the relevant distributions up to estimable parameter values. Unfortunately, there are few topics in economics where these assumptions are tenable, at least in full. This chapter discusses path models and their use in the modeling of rudimentary theories. It also discusses the partial least squares (PLS) approach to path models with latent variables.