A statistical inference comparison for measurement estimation: Application to the estimation of groove dimensions by RFEC

Abstract

The purpose of the current paper is to present the comparison of different techniques for making statistical inference about a measurement systemmodel. This comparison presents results when two main assumptions are made. First, the unknowable behavior of the errors probability density function (pdf) p(e), since the real measurement systems are always exposed to continuous perturbations of an unknown nature; second, the assumption that after some experimentation one can obtain suf cient information which can be incorporated into the modelling as prior information. The first assumption lead us to propose the use of two approaches which permit building hybrid algorithms; such approaches are the non-parametric Bootstrap and the kernel methods. The second assumption makes possible the exploration of a Bayesian framework solution and the Monte Carlo Márkov Chain (MCMC) auxiliary use to access the a posteriori measurement pdf For both assumptions over p(e) and the model, different classical criteria can be used; one uses also an extension of a recent criterion of entropy minimization. Finally, a comparison between results obtained with the different schemes proposed in [9] is presented.