Skip to main content
. 2020 May 20;23(1):207–222. doi: 10.1093/biostatistics/kxaa015

Fig. 2.

Fig. 2.

The log-ratio model described in Section 3 can only be fit to data with a minimum abundance greater than zero. Abundance data for microbiome studies are generally sparse, and 42% of the observed abundances of the Lee and others (2015) dataset are zero. For this reason, it is common to add a perturbation offset Inline graphic to the observed abundance table before fitting the log-ratio model. Here, we see that the estimated diversity does depend on the choice of Inline graphic.