Model Selection for Investigation of the Field Distribution in a Reverberation Chamber

In this work two model selection criteria, i.e., Akaike's information criterion (AIC) and minimum description length (MDL), are applied to measurements in a RC with Rayleigh, Rician, Nakagami, Bessel K, and Weibull distributions as the distribution candidate set. In spite of small differences of the AIC and MDL tests (due to their different penalty terms on distribution parameters), both criteria result in similar conclusions. Results show that the Rayleigh distribution provides the overall good fit to the Cartesian field amplitude, especially for an overmoded RC, and that the Weibull distribution provides good fit to the Cartesian field amplitude in an undermoded or loaded RC. In addition, it is found that both the Rician and Weibull distributions provide improved approximations of the Cartesian field amplitude in a loaded RC with non-negligible unstirred components and that the transition from undermoded RC to overmoded RC depends not only on the operating frequency and mode-stirrer efficiency (as it is commonly believed) but also on source stirring and RC loading.