The following is based on the studies by MacLeod et al. (2002) and Slob (1994) on this topic. A simple sensitivity analysis is done first in which a test is made of the magnitude and sign of the sensitivity of each output parameter (y) to each input parameter (x). This yields an x–y sensitivity or S-matrix, each entry of which is the partial derivative of each value of ln(y) to each ln(x):