Pivoting for Size and Sparsity in Linear Programming Inversion Routes

^* Present address: The Operations Research Department, Stanford University, Stanford, California 94305, U.S.A. This note reports some experimental results on the inversionof real linear programming bases, with particular emphasis onthe compromise between minimum density and maximum numericalstability. The general features of a linear programming inversionroutine are outlined and the special structure of linear programsconsidered. The main result is a suitable and apparently safe"pivot tolerance" level, together with more general data onthe nature and behaviour of the problems.