Rational and polynomial lags: The finite connection

This article demonstrates that, for a finite distributed lag, the polynomial distributed lag (PDL) approximation suggested by Almon is a special case of the rational lag method formalized by Jorgenson. The proof relies upon the fact that the PDL estimator imposes differencing restrictions upon the parameters while rational lag methods impose quasi-differncing restrictions. Because of this relationship, the PDL restrictions are nested inside the rational lag ones, and this provides for a sequence of tests to discriminate between the two. An example is performed and an appendix describes an asymptotically efficient two-step estimator.     

A comparative analysis of multiattribute attitude models

In the consumer behavior context, multiattribute attitude models used to predict consumer choice in multiple criteria decision making situations have produced mixed results. Prediction of consumer behavior from attitudes, consequently, has been highly debated in recent years with researchers using beliefs-only, full, and extended multiattribute attitude models. The research underlying this paper was designed to compare the predictive superiority of the beliefs-only model, the full multiattribute attitude model, and a new representation identified as the combined multiattribute/determinant attribute attitude model. Data concerning patronage of fast-food chains were collected from a student panel for seven weeks. Predictive superiority was determined by average adjusted R² using the patronage behavior dependent variable. All models were tested in aggregated and disaggregated form.