| 摘要 |
Disclosed is a computer implement statistical technique for normalizing the response curves of multiple measurement methods. The end results is a group of response curves, one for each measurement method under consideration, which depend on a common independent variable -- the actual physical property being measured by the methods. The results are provided as a collection of equations, curves, and/or tables (a "nomogram") to facilitate conversion of measured values from one method to measured values from a second method. In the technique, data is provided for each of the methods being normalized. The input data includes measured values from common samples which are analyzed by two or more of the methods under consideration. The technique also requires assumptions or approximations of the true physical property values for each of the samples used to generate the data. Still further, the technique requires assumptions of the mathematical form of the response curves (e.g., linear, sigmoidal, etc.). The computer system then solves all parameters appearing in the mathematical expressions simultaneously. At the same time, the computer system solves for one or more "correction factors" to some of the approximations for the true physical property values of the samples. This specifies the relative separation distances between the physical property values on the independent variable axis. |
