| 摘要 |
PROBLEM TO BE SOLVED: To suppress user's burden smaller, while estimating uncertainty of a measurement system difficult to analyze theoretically by a Monte Carlo method, and to enable the estimation of the uncertainty with high reliability. SOLUTION: Measurement values, obtained from a measuring device of object to be estimated, is inputted, and errors in measurement values is estimated (S1). A covariant matrix or a correlation matrix in measurement values are determined (S2), based on the estimated errors in measured values. Characteristic values and eigenvectors are determined, by decomposing the covariance matrix and correlation matrix into characteristic values (S3). The pseudo-measurement values of the measuring device are produced, by producing a normal random number that expected values to each eigenvector are 0 and dispersion is a characteristic value corresponding to an eigenvector, as a coupling coefficient (S4). The uncertainty of the measuring device is estimated (S5), by statistically treating the obtained the pseudo-measurement values. COPYRIGHT: (C)2006,JPO&NCIPI
|
