On evaluation and comparison of measurement methods with special reference to the meat and meat products

Abstract

The authors discuss the main systematizing principles, not generally used in the field of meat science, regarding critical appreciation and comparison of measurement methods. The presentation of some basic considerations is followed by the discussion of interfering factors. Besides the random error, the role and importance of the constant, proportional and random biases are also treated. The random error is related to the method, while the random bias is related to the test material. A clear distinction is necessary between deterministic and stochastic relations when various methods are compared, emphasizing a possible instability of the regression constants in the latter case. A number of examples are listed for both types of relation. The use of the correlation coefficient may be erroneous in comparing two methods. However, Mandel’s sensitivity ratio often furnishes an appropriate basis of comparison. For better understanding, the considerations on biases are illustrated by a simplified example. Thereafter, the authors deal with the most important view-points in developing and revising a method of measurement: simplicity, robustness and instability. The handling of extreme observations or outliers and the concentration dependence of the relative standard deviation (coefficient of variation) of measurement are also discussed.

Determination of fat content in meat by pulsed nuclear magnetic resonance (p-nmr) technique

“…Results are presented in Tables 1 and 2. The evaluation of data (KÖRMENDY et al, 1989;KÖRMENDY & ZUKÁL, 2002) showed that the relations presented in Table 1, except sample with serial number 5, can be described with the y ˆ = b.x equation, i.e. proportional relations exist between intensity of P-NMR signal (y) and fat content (x) measured with the Soxhlet method.…”