Ein statistisches Verfahren zur Entdeckung von Aenderungen der systematischen Abweichung bei Messreihen

Analytical measurements are associated with uncertainties which in many cases can be characterized by two parameters. These two parameters are the standard deviations of the random deviations and of the biases. Random deviations are understood to mean those which undergo random variations from one single measurement to the following measurement on account of accidental superposition of a more or less great number of influences by disturbing variables changing with the time, although all experimental conditions are kept constant. The biases are those which do not undergo variations under the said conditions. Besides, it has repeatedly appeared that the assumption of constancy of biases over rather long periods of time can actually not be made in all cases. Rather it is reasonable to suppose that occasionally abrupt jumps may occur in the bias (caused e.g. by calibration of the measurement instrument). These jumps in the experimental boundary conditions may exert a major influence on the evaluation of the measured data. In this study a novel statistical method is described which allows to determine whether and, if so, when changes in the bias occur in a measurement series to be investigated. In addition to description of the method, detailed information is provided on its reliability. The method was tested by concentration assays of chemical elements and to both simulated data and data measured in industrial scale manufacture. It is demonstrated that the method is capable of making a valuable contribution in ''Process and Quality Control''. (orig.)Available from TIB Hannover: ZA 5141(5573) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekSIGLEDEGerman