Industrial corporations frequently require that different analytical facilities or instruments he compared with one another. Traditionally, the isoplot technique is used to accomplish this goal. This technique is only useful, however, when each subject has analyzed a large number of identical samples. Because this is not always the case, the present paper presents an alternate method of comparing analytical results. It is designed for cases where subjects have analyzed a relatively small number of trials from samples that are not identical. The proposed method uses normalized scores and ordinary least squares (OLS) linear regression. In the present case, the proposed method was applied to a situation where the research team compared five different optical emission spectroscopy (OES) laboratories that had tested different sets of OES calibration standards. The normalized scores method provided a quantitative comparison of the laboratories in question. Based on this analysis, it was possible to conclude that one lab (B) clearly lagged behind the others, with respect to the precision of their analyses. It was also apparent that none of the labs were completely accurate. Each either overestimated or underestimated the calibrated element levels.
展开▼
