should be 3

(Not essential)

This and the other barplot are ok, but here, because most datasets are so dominated by human reads, I'd try other representations.

I also don't really like aggregating all samples together like this, since it masks so much variation among samples, and would personally consider ways to plot the variation across samples.

this is problematic; it can create a serious bias. You really want to find a way to include these samples, even if it makes the boxplot extend to the axis limit.

plot suggestions * add an x-axis label * increase the separation between the plots

I suggest using round() or signif() (or another formatting function) to simplify the numbers in the table

I think this should be 'Phase 3'

Correct link: https://doi.org/10.5281/zenodo.5706553

Note, this use of 'proportion' differs from that earlier in the article

List currently in the wiki at https://github.com/mikemc/differential-abundance-theory/wiki/Empirical-methods-for-differential-absolute-abundance-inference

This sentence misidentifies type I error: A type I error is when the null hypothesis is incorrectly rejected (see e.g. https://en.wikipedia.org/wiki/Type_I_and_type_II_errors#Type_I_error)

Note, the data shown for Figure 1B days 7 and 14 is identical

Based on the FACS abundances supplied in the SI materials and the metagenomics abundances as determined by Metaphlan2, I determined that the three species V. cholerae, C. saccharolyticum, and Y. pseudotuberculosis in this figure are mislabeled. See the methods section of https://doi.org/10.7554/eLife.46923 and my attempt to reproduce this figure in https://mikemc.github.io/mgs-bias-manuscript/analysis/costea2017-analysis.html#reproduce_costea_et_al%E2%80%99s_figure_6a

The right-hand side should be multiplied by 1/2 to account for the missing factor of 2 in the derivation; see my note above

A factor of 2 is missing from the right-hand side of this equation, which results in the expression for the variance in Theorem 5 being 2 times too large.