Answers are grounded in this paper and cite their sources.
A non-parametric bootstrapping statistical method is introduced and investigated for estimating confidence intervals resulting from information transfer (IT) analysis of confusion matrices. Confidence intervals can be used to statistically compare ITs from two or more confusion matrices obtained in an experiment. Information transfer is a nonlinear analysis and does not satisfy many of the assumptions of a parametric method. The bootstrapping method accurately estimated IT confidence intervals as long as the confusion matrices contained a sufficiently large number of presentations per stimulus category, which is also a condition for reduced bias in IT analysis.
No takes yet. Share an insight, caveat, or question.
Azadpour et al. (2014) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: