Currently, I have LazyData: true in the DESCRIPTION file, which means that the datasets in data/ are available without calling the data() function. If I keep this setting, then I can remove the calls to data() from the Rmd's.

Calculating the error with adist as in

error <- pred %>%
    group_by(Mixture_type, Bias_type) %>%
    summarize(
        RAdist = sqrt(mean(adist(Observed, Predicted))),
        Adist2 = mean(adist(Observed, Predicted)^2),
(...)
        )

is incorrect, as the Aitchison distance must be calculated within samples and then averaged. This is currently done twice in the Brooks2015 analysis, though the results are not currently used or referenced in the manuscript.

Possible ways to improve documentation include adding a README.md to the folder giving the overall purpose and a brief description of each script, and adding a numerical prefix to the scripts for the costea2017 dataset by order in which they need to be run.

In particular, the estimated CN of 1 for L. iners is suspicious. Should compare the estimates from the refseq genomes to the closest relatives in the rrnDB, and if warranted, use the median of the rrnDB numbers instead of the refseq annotation.

also make sure coord_fixed() is used

Specifically in case the new nest() and unnest() syntax breaks the existing code. (See https://tidyr.tidyverse.org/articles/in-packages.html.) If so, either update the code or create a requirement for tidyr < 1.0.0 and a warning in the ReadMe

Many thanks for providing the tools for using correcting bias in metagenomic sequencing measurements. I still need to fully understood the paper but it is very interesting.

However I couldn't use the method implemented on the paper by calling the package with library. The package only exports some data: "%>%", "brooks2015_counts", "brooks2015_sample_data", "brooks2015_species_info", "costea2017_metaphlan2_profiles", "costea2017_mock_composition", "costea2017_sample_data".
To correct this you need to add the @export tag on each function that you want to make available for the user. I could provide a pull request if you wanted.

Currently the figures make heavy use of ggplot's default colors for a three-level categorical variable---red, green, and blue---making essential variables impossible to distinguish for people who are red/green colorblind. To fix, need to pick new colors for the taxa in Figures 1 and 2, and change the text references to e.g. "the red taxon", and also need to pick new colors to distinguish the three protocols in the shotgun dataset.

Just a note to archive a copy of this repository somewhere more permanent (e.g. Zenodo) before publication. Can use Zenodo's Github integration https://zenodo.org/account/settings/github/

Cowplot is back to overriding the default ggplot theme, and this is messing up the sizing in some figures in the html documents

in the Brooks2015 dataset.

Using the simple linear model of Krehenwinkel2017 and the all-taxa three-way interactions model of Brooks2015. Add to the brooks2015 analysis and save for SI figures.