Analysis Of Electromagnetic Susceptibility Data

Firstly, this paper describes the Logistic regression theory which can be applied to the electromagnetic susceptibility data. Comparing the maximum likelihood estimation method and the weighted least squares method, we learn that the maximum likelihood estimation method needs to know the distribution of samples , and the weighted least squares method relies on data packets. And we introduce goodness-of-fit tests of model and give three evaluation methods: deviation test, Pearson test and Hosmer-Lemeshow goodness-of-fit indicators; then introduce the significant tests of the regression model and coefficients and the confidence intervals of the regression coefficients and the event probability, also introduce the two types of problems which can affect the parameter estimation.Secondly, we give in-depth discussions of the data structure problems and introduce the variable selection criteria and model verification methods of the logistic regression model, and these methods are very useful for processing the electromagnetic susceptibility data in the complex electromagnetic environments. Because the factors, which affect the electromagnetic susceptibility of the electronic equipments in the complex electromagnetic environments, are not only the results from the theories, but also some results existing in the surrounding electromagnetic environments. Therefore, discussions of the data structure problems and the variables selection of the model and model verification before building it, will help us to establish more effective forecasting model.Thirdly, this paper analyse electromagnetic susceptibility data of the incident test of single-point source in the hole-infinite flat plate. We mainly study the relationships between R (the distance between point source and the aperture), d (the vertical distance between point source and the infinite flat plate) and the response in the observation point. Then we calculate the corresponding simulation data, and establish the logistic regression model; and using the simulation model test data, we verify the validity of the model. Next, using statistical analysis theory, we estimate the model parameters, and test the goodness-of-fit and the test results show that the model fit the data well. Also, we give the confidence interval of the regression coefficient and the event probability. When the confidence level is 95%, the confidence intervals of regression coefficients are (-4.5533, -2.3891) and (-2.3891, -0.5060) and the confidence interval of the event probability is (0.0455, 0.1318), which show that the model has good credibility. Finally, we research the prediction standards and forecasting methods on electromagnetic susceptibility data, and use the interpolation and extrapolation methods to verify the effectivity of the logistic regression forecasts, that is, in which location logistic regression forecast is effective and in which position it needs for further verification. Next, we will use logistic regression to analyse multi-point source, or more complex electromagnetic susceptibility tests.