The LP Procedure

Scaling

Based on the SCALE= option specified, the procedure scales the coefficients of both the constraints and objective rows before iterating. This technique can improve the numerical stability of an ill-conditioned problem. If you want to modify the default matrix scaling used, which is SCALE=BOTH, use the SCALE=COLUMN, SCALE=ROW, or SCALE=NONE option in the PROC LP statement. If SCALE=BOTH, the matrix coefficients are scaled so that the largest element in absolute value in each row or column equals 1. They are scaled by columns first and then by rows. If SCALE=COLUMN (ROW), the matrix coefficients are scaled so that the largest element in absolute value in each column (row) equals 1. If SCALE=NONE, no scaling is performed.