Fig. 2.

Empirical power based on 1000 simulated datasets with n = 100, two time points and continuous outcomes. K$\text{omni}\hspace{0.5em}$ is the omnibus test, K₁ and $K_{0.5}$ are generalized PUniFrac with γ  =  1 or 0.5, K_U is unweighted PUniFrac, K_JQ is the quantitative paired Jaccard kernel, K_JB is the qualitative paired Jaccard kernel, and K_JC is the Jaccard kernel using CLR-transformed proportions. The continuous outcome is associated with: (A) Change in presence of a rare cluster. (B) Change in presence of 60 randomly selected taxa. (C) Change in abundance of a moderately common cluster. (D) Change in abundance of the 10 most abundant taxa. (E) Change in CLR-transformed abundance of a moderately common cluster. (F) Change in CLR-transformed abundance of the 10 most abundant taxa