Uncertainty Evaluation in Multivariate Analysis – A Test Case Study

摘要

We have used different multivariate analysis methods to estimate quantities in the fields of food control and atmospheric remote sensing. In order to estimate the uncertainties in these estimates we studied analytical as well as non-parametric numerical methods. The methods have been evaluated by comparison between obtained results and independent sets of measurements. We present one test case from each field, including results, where these methods have been applied. For the food control test case reduced chi-squared of approximately unity indicate that both the analytical and numerical methods used for uncertainty estimation produce uncertainties of reasonable size. In the atmospheric remote sensing test case, a indicated that the uncertainties from the numerical method were far too small, whereas a indicate that the size of the analytically determined uncertainties can represent the size of the “true” errors.