Calibration when using new methods of analysis

When preparing a method of analysis, extensive calibration should be 

The latter condition means that standard solutions must be taken not from a previous step in the analysis but from separate ones. Alternatively, if this is not possible, independent dilutions from a single parent solution must be prepared. Dilutions made in succession lead to systematic errors. 

In the preparation and familiarization stages of a new analytical method, DIN 38 02 recommends that the information values y be established for N = 10 calibration samples. In so doing, the intervals of the individual concentration levels should be equidistant, and so equally distributed over the working range. In order to verify that the variance of the information values Y[ is independent of the concentration, a statistical test of the homogeneity of the variance can be used. For this purpose one analyses n = 10 standard samples of the lowest (yj,l) and n = 10 standard samples of the highest concentration level (yj io)  

The homogeneity of the variances is checked using the F-test, for which the test quantity PG is determined by 

Apart from homogeneity of the variance, the linearity test, as already mentioned, is also a precondition for being able to use linear regression as a calibration function. 

---