KL divergence is a well known asymmetric measure of distance between distributions, but when densities are sampled instead of computed, distances and differences may be obfuscated by sampling variability and lack of certainty about a reference density. This work showcases a model for assessing whether or not densities are statistically distinguishable using tools from functional data analysis with application to assessing if two or more MCMC runs have converged to the same limiting distribution. The problem is formulated using functional generalized linear models to classify if samples came from a particular MCMC run. The model is interpretable as a symmetric measure of disagreement between sampled densities but rather than an integrated measure, it showcases specific regions and directions of distributional dispute.