The use of probability density functions to improve the interpretation of FSV results

Abstract

This paper presents research undertaken using a continuous probability density function as a supplement to the confidence histograms in Feature Selective Validation. Such an approach shares the benefits of `categorisation' from a humanuser perspective but, because of the continuous nature of the distribution, it also allows further inter-comparison analysis to be undertaken such as using non-parametric tests on the distributions. This approach improves the analysis of FSV data when multiple comparisons are required to be performed.