[Task] Joss paper

[smasongarrison]

finish up joss paper

[smasongarrison]

nearly done :) https://joss.theoj.org/papers/e5116b83b03e2740960d1153c45f9480

[smasongarrison]

[smasongarrison]

openjournals/joss-reviews#8624

[smasongarrison]

openjournals/joss-reviews#9434

[smasongarrison]

To do list

• sex column required? #92 Added sex column requirement
• [REVIEW]: ggpedigree: Visualizing Pedigrees with 'ggplot2' and 'plotly' openjournals/joss-reviews#9434 (comment)

I've now read the paper and played around with the package, overall a pleasant experience. The package provides highly customizable pedigree plots based on the ggplot2 framework, and also plotly versions with some degree of interactivity. Although alternatives for this already exist (e.g. https://louislenezet.github.io/Pedixplorer/) the approach here is quite different in practice, with a more a ggplot2-centric workflow for customization.

I did have some initial trouble using the package, partly due to insufficient documentation, I've reported this in separate issues in the ggpedigree repo. Below are general comments on the manuscript and the software.

• Unexpected behaviour with simple pedigrees #95 Unexpected behavior
• Default plot appearance #94 Default Plot
• Singleton plot #93 Singleton plot

• Authorship: The package includes a substantial amount of code copied from kinship2, since that package may be retired. I am unsure what the best way to handle this is, but I would at least expect the DESCRIPTION to acknowledge it and give appropriate credit for the borrowed code. As it stands, the DESCRIPTION presents ggpedigree as an alternative to kinship2, which seems only partly true given that much of the plotting logic is inherited from kinship2.

• Update description 1737bc8 fc92a58

• The paper acknowledges that the package adapts algorithms from kinship2, "with enhancements for large, multi-generational pedigrees and complex family structures". This is interesting and deserves more explanation. As kinship2 is the currrent R standard for aligning large & complex pedigrees, such claimed improvements should be clearly described and ideally accompanied by runtime comparisons or similar.

• Bug Co-author about algorithm revision at BG

• On the same note, the online vignette https://r-computing-lab.github.io/ggpedigree/articles/v22_plots_morecomplexity.html#visualize-the-pedigree-with-ggpedigree shows a large-ish pedigree plotted by both kinship2 and ggpedigree. The example is surprising, since the ggpedigree versions are visually inferior to the kinship2 one. I would recommend reworking this example so that it better highlights the strengths of ggpedigree.

• You're right. I've reworked it so that now it looks nicer than the kinship2 example.

• The paper presents ggpedigree as a stand-alone package for pedigree plotting, but in reality it is intertwined with another package, BGmisc for analysis in behavioural genetics. While this may work fine in principle, the separation between the two feels a bit ad hoc and incomplete. I think the project as a whole would benefit from clarifying the responsibilities of the two packages. A free-standing ggplot2-based pedigree plotter would likely appeal to users across many fields, but conceivably less so if it is too closely tied to behaviour genetics applications.

• address the dependency logic. make it clear why bgmisc is required for the relatedness. But that's a fair point about overfocusing on the behavior genetics angle. I've now downplayed the BG connection in the paper as well as the description.

• The package includes two built-in example pedigrees, redsquirrels and ASOIAF, but both give errors when I try to plot them, and their help pages don’t show any example code. There are working examples elsewhere, but they rely on datasets from BGmisc. For a single-purpose package I would expect simple, self-contained examples.

• add simple examples to show how to use in documentation, aeb7661
• as well as direct to relevant vignettes

• Several interesting exported functions, like calculateCoordinates(), lack examples in their help pages. Please add simple, self-contained examples. (I actually thought examples was a CRAN requirement, but apparently not?)

• add

• Many exported functions of the package have very similar names:

  • ggPedigree
  • ggpedigree
  • ggPedigreeInteractive
  • ggPedigreeinteractive
  • ggpedigreeInteractive
  • ggpedigreeinteractive
  • pedigree
  • plotPedigree

  I understand that many of these are simple aliases, but in my view this has more cons than pros: It clutters tab-completion, makes the documentation hard to navigate, and obscures which function is intended as the main one. For example, the fact that plotPedigree is not the primary plotting function is quite confusing to me. I recommend to simplify the API and reduce some of the redundant functions.

• fixing by deprecating several of them as well as the plotPedigree wrapper aaa08b8

• The package imports reshape2 alongside tidyverse package, which seems unneccessary. As I understand it, the authors of reshape2 have recommended switching to tidyr in new code.

• removed dependency cd4ad5a

[smasongarrison]

Update cite on readme