Hi (to all). There are some theory for situation like this (and, fortunatelly, functions too): condition numbers . Those are the mathematical equivalent for the expresion "bad conditioned problem" (or ill, or well). In mathcad you have cond1, cond2, condi and conde, based on different norms, and obviously are related with some determinants near zero, and rows or columns which are reduced by rref. Remember that a det = 0 implies loosing a degree of freedom, and a row n rref function. Thats equivalent to say that the vectorial space defined by the columns have not the dimension of the columns, because the "vectors" (rows) are not lineal independent.
Anyway, cond numbers reflect those and some other issues in input data. Inroducing noise in data can be a solution, but can justify completely using those indicators.
Best regards.
Alvaro.
