Do starting values really matter? Development of a genetic algorithm approach

Abstract

Evaluating the impact of using different starting conditions in econometric models is increasingly critical. This study demonstrates implications of using alternative starting value conditions and how these implications vary with econometric model complexity. This piece also proposes a universal approach in developing appropriate starting values for use in subsequent model estimation.

Notes

¹ Our discussion on GA is only a ‘sufficient introduction’. Those further interested are encouraged to consult the Dorsey and Mayer article.