Bayesian method of inversion and a new strategy of inference on parameter estimation ? An application on MT data

Abstract

We describe a new interpretation technique in geophysical inversion that uses Box-Jenkins schema for model selection using Bayesian paradigm while model parameters are estimated using maximum a posteriori estimation procedure. We employ a regularized Gauss-Newton iterative optimization technique for selecting model parameters. Model discrimination and parameter appraisal have also been realized in the Bayesian framework. The model adequacy is tested using Bayesian information criterion (BIC) and logarithmic evidence criterion (LEC). The method is demonstrated with an application to field data.