Supplementary Fig. 11 shows how excluding the non-credible samples changes the 95% HDI for the example of healthy vs. non-inflamed samples of ulcerative colitis from the Lamina Propria1. While excluding the zero samples from the HDI calculation influences the HDI of most cell types only marginally, some high-density intervals become slightly wider (CD69- mast cells) or shift away from zero (cycling B cells). The average width of 95% HDIs increases only slightly from 0.92 to 0.97, though. Note that generally Bayesian high-density intervals are relatively large due to the MCMC sampling uncertainty.
