Calibration and validation of on-line measurements Analysis of results

Instrument test data are analysed to determine whether any adjustment is needed, if this adjustment affects zero shift only or gain and zero shift, and the magnitude of any adjustment needed. This paper uses practical data to demonstrate calibration hunting whereby successive tests require gain to be alternately increased and decreased, and tests for skew and bias. Other sections deal with confidence intervals, rejection of outlier results, regression methods, and the correlation coefficient.