Multinomial Regression Model for the Evaluation of Multilevel Effects Caused by High-Power Electromagnetic Environments.

A lot of work has been done for the effects evaluation of electronic equipment due to high-power electromagnetic environments. The focus of the evaluation usually stays on whether the effects occur or not (“1” or “0”) during tests. However, in addition to such kind of either-or effects, multilevel effects happen in many practical cases as well. In this paper, we propose a kind of statistical approach for the evaluation of multilevel effects. By means of multinomial regression, the occurrence probabilities of each level of effects can be estimated simultaneously. Through maximum likelihood estimation (MLE), both the Softmax model and the mutually independent normal distribution assumption model are built and solved in a numerical way. We also talk about how the data affect the goodness of the regression. As a consequence, the Monte Carlo simulation gives a quasi-function relationship about model stability and estimation error. A case study on computer communication system has been conducted in the laboratory to assess the validity and applicability of the proposed models. The result is satisfactory for the three-level cases and gives all the probability prediction of the malfunction occurrences. [ABSTRACT FROM AUTHOR]