Updating methods for adaptive econometric models

Abstract

"Objectives: This study examines the problem of updating the structure of simultaneous equations models of the agricultural sector. The topic is specialized as updating is assumed to take place in the structure. In light of the empirical rationale established in the previous section, it is also assumed that updating is based on new sample observations and/or alternative parameter hypotheses. Updating with alternative parameter hypotheses is viewed as a method for incor­ porating the different values the structural coefficients may take over time, i.e., parameters that may follow an evolutionary structure. Time-varying parameters or evolutionary parameters hypotheses and new sample data are attractive sources of information for updating as their consequences have a natural interpretation. The specific objectives of the study are five-fold: (i) Characterize the problem of updating with variational parameters hypotheses and additional sample observations. Concern is given to the information source, information value, and the restrictions implied on the parameters. (ii) Develop efficient estimation methods for updating with variational parameters hypotheses and additional sample observations. Interest focuses on those methods that can produce estimators with statistically desirable properties. Computational efficiency is also an important characteristic to be studied. (iii) Generalize the updating problem for the case when the two sources of information, variational parameters hypotheses and additional sample observations, are merged. Develop estimation methods for accommodating updating based upon the two information sources in a simultaneous equation framework. (iv) Illustrate the process of updating in an applied context. Examine a provisional representation of an econometric model and apply alternatively justified and designed updating methods in moving to an "improved" representation. (v) Evaluate the results of the applied updating case. Assess the implications of the incorporated additional information and develop guidelines for use in large scale modeling exercises for the agricultural sector."--Page 3.