Penalty calculations and branching rules in a LAV best subset procedure

Least absolute value (LAV) regression has become a widely accepted alternative to least squares regression. This has come about as the result of advancements in statistical theory and computational procedures to obtain LAV estimates. Computer codes are currently available to solve a wide range of LAV problems including the best subset regression. The purpose of this article is to study the use of penalty calculations and other branching rules in developing the solution tree for the best subset LAV regression.