Assessment of description quality of models by information theoretical criteria based on Akaike and Schwarz-Bayes applied with stability data of energetic materials

To describe any measurement data by models or parametric equations is mostly a necessity for the interpretation and further evaluations of the data. There is often the case that several models may be applicable and the question arises, which model is the better one. Besides many criteria to argue for one or another model there are objective methods for the assessment of the description quality for models in comparison. Such methods are based on information theoretical conclusions and two criteria have proven helpful. One has been developed by Hirot(s)ugu Akaike and it is called Akaike Information Criterion (AIC). The other method is based on the early work of Thomas Bayes, which was adapted by Gideon Schwarz to the framework of an information criterion and is named Bayes Information Criterion (BIC). With several data sets obtained with energetic materials and some models applied on them the usefulness but also limitations of these two information criteria are demonstrated and discussed. The data comprise stabilizer consumption and molar mass degradation of nitrocellulose in a gun propellant as well as the degradation of cellulose in electrical transformer insulation paper.