Korun M, Maver Modec P, Vodenik B
Jožef Stefan Institute, Ljubljana, Slovenia.
Appl Radiat Isot. 2012 Jun;70(6):999-1004. doi: 10.1016/j.apradiso.2012.03.013. Epub 2012 Mar 18.
Empirical evidence is provided that the areas of peaks having a relative uncertainty in excess of 30% are overestimated. This systematic influence is of a statistical nature and originates in way the peak-analyzing routine recognizes the small peaks. It is not easy to detect this influence since it is smaller than the peak-area uncertainty. However, the systematic influence can be revealed in repeated measurements under the same experimental conditions, e.g., in background measurements. To evaluate the systematic influence, background measurements were analyzed with the peak-analyzing procedure described by Korun et al. (2008). The magnitude of the influence depends on the relative uncertainty of the peak area and may amount, in the conditions used in the peak analysis, to a factor of 5 at relative uncertainties exceeding 60%. From the measurements, the probability for type-II errors, as a function of the relative uncertainty of the peak area, was extracted. This probability is near zero below an uncertainty of 30% and rises to 90% at uncertainties exceeding 50%.
有经验证据表明，相对不确定度超过30%的峰面积被高估了。这种系统影响具有统计性质，源于峰分析程序识别小峰的方式。由于这种影响比峰面积不确定度小，所以不容易检测到。然而，在相同实验条件下的重复测量中，例如在背景测量中，可以揭示这种系统影响。为了评估这种系统影响，使用Korun等人（2008年）描述的峰分析程序对背景测量进行了分析。影响的大小取决于峰面积的相对不确定度，在峰分析所用的条件下，相对不确定度超过60%时，影响可能达到5倍。从测量中提取了II类错误的概率，作为峰面积相对不确定度的函数。在不确定度低于30%时，该概率接近零，在不确定度超过50%时上升到90%。
