Selecting the correct model to fit data, especially with impact on critical decisions, requires careful control of the goodness of fit assessment to guarantee a high reliability. One example is the modelling of radioactive contamination as a function of geographic location, which is relevant for radiation protection decisions and missions of mobile incident monitoring. IMSD has developed such a tool recently on behalf of the German radiation protection service.
In this talk, we present the goodness of fit assessment technique of our tool, allowing a) to visualize and assess the chi-squared via its decomposition by degrees of freedom and b) to resolve any weaknesses of chi-squared terms via model deformations, linear-logarithmic transformations and other techniques in compatibility with theoretical constraints. We demonstrate our technique for the
example of radioactive contamination.
Time: 14:30 - 15:00
organised by SSS
Speaker: Pierre Mandrin