Using Akaike Information Criterion for Selecting the Field Distribution in a Reverberation Chamber

Journal article, 2013

Previous studies on modeling the random field (amplitude) in a reverberation chamber (RC) were conducted either by fitting a given distribution to measured data or by comparing different distributions using goodness-of-fit (GOF) tests. However, the GOF tests are inappropriate for comparing different distribution candidates in that they are meant to check if a given distribution provides an adequate fit for a set of data or not and they cannot provide correct relative fitness between different candidate distributions in general. A fair comparison of different distributions in modeling the RC field is missing in the literature. In this paper, Akaike's information criterion (AIC), which allows fair comparisons of different distributions, is introduced. With Rayleigh, Rician, Nakagami, Bessel K, and Weibull distributions as the candidate set, the AIC approach is applied to measured data in an RC. Results show that the Weibull distribution provides the best fit to the field in an undermoded RC and that the Rayleigh distribution provides the best approximation of the field in an overmoded RC. In addition, it is found that both the Rician and Weibull distributions provide improved approximations of the field in an RC loaded with lossy objects. This study provides correct complementary results to the previous RC studies.