The MGA overcomes this problem by essentially fitting a line between two subsamples: one consisting of dependent variable values (forward price) that fall below the median and another consisting of independent variable values that fall above the median