Comparison of analytical methods

A technique for comparing two analytical methods is presented in this paper. Frequently a new method, generally simpler, faster or less expensive, is compared with a reference method. The deterministic relationships between two methods, subject only to random errors, is first discussed. However, the relationship between two methods may be stochastic (non-deterministic); this problem is also discussed here. The concept of the sensitivity ratio of Mandel furnishes an appropriate basis of comparison. The use of the correlation coefficient may be completely erroneous for this purpose. Tests are presented for the significance of constants in the regression equation. Dangers in the use of stochastic relationships are strongly emphasized because of the unstable estimation of constants in the regression formula. Some systematizing principles not generally known by food scientists are presented here.