Fig 1.
A diagram of statistical methods used in the study.
Table 1.
An overview of statistical methods and their properties.
Fig 2.
Explained variance for simulated data and the relative number of simulations where the simulated effect was detected.
Numbers indicate the percentage of simulated data sets where p-value was significant.
Table 2.
Community-level method comparison across five experimental data sets.
Fig 3.
Sensitivity (True Positive Rate) for the four scenarios in the simulation study.
Fig 4.
Spearman’s correlation (Y-axis) calculated for pairwise comparison of statistical methods (X-axis) for (A) simulated data and (B) five experimental data sets.
Each point represents Spearman’s rank correlation coefficient between OTU ranking metrics from the two methods compared.
Fig 5.
Mean relative abundance (log-scale) plotted versus ANCOM W-stat.
(A) The “Few-Low” simulation scenario and (B) Birkeland data set, (C) “Many-Low” simulation scenario and (D) Moen data set. Red points (panels A and C) indicate differentially abundant OTUs.