Fig. 3.

Bootstrap sampling of simulated distance matrices allows PERMANOVA power estimation. An example of PERMANOVA power estimation by bootstrap sampling from simulated distance matrices is shown. The horizontal axis depicts the effect sizes simulated by segregating OTU group membership, and the vertical axis represents the power to detect the effect, as determined by the proportion of bootstrap distance matrices drawn from the simulated distance matrix for which PERMANOVA P-values were below the specified type I error threshold. The simulated study includes three exposure groups; the power to detect effects with five, 10 and 20 subjects per group is depicted in red, green and blue, respectively. Power estimated for a null effect (i.e. differences between subjects but not between groups) was equal to PERMANOVA type I error