search

hubverse-org / hubEnsemblesManuscript

https://htmlpreview.github.io/?https://github.com/Infectious-Disease-Modeling-Hubs/hubEnsemblesManuscript/blob/master/analysis/paper/hubEnsembles_manuscript.html
Other
1 stars 2 forks source link

point for discussion: how much code do we really need to show? #57

Closed elray1 closed 3 months ago

elray1 commented 3 months ago

I recognize that I'm asking us to re-litigate old discussions. But my feeling is that at some points in this manuscript, we're showing a lot of code that's essentially uninteresting and is not the point of the manuscript. The purpose of this paper (and vignette) is ultimately to illustrate hubEnsembles functionality, and I feel that this code is getting in the way. As an example, I'm looking at the code for creating Figure 5, which occupies nearly 2 pages and contains no code related to computing ensembles. My vote is that we could spare the reader this stuff. In a package vignette, if possible I might say we could leave the code in but hide it if we can figure out how to do that easily (some "click to expand" kind of thing). But in an article for Journal of Statistical Software, it kind of just feels inappropriate to me -- I would prefer to suppress it, and leave in a more limited/intentional selection of code. Maybe the code that shows what the input and output data frames are and computes the ensembles. And one or two hubVis calls as appropriate?

elray1 commented 3 months ago

I feel similarly about the code for figure 6: this is a lot of space to devote to some code that is not really related to the main point of the paper.

annakrystalli commented 3 months ago

Click to expand available for HTML documents via code block option code-fold. More details here: https://quarto.org/docs/reference/formats/html.html#code

nickreich commented 3 months ago

I think I agree with Evan about this. Maybe a middle ground is for the static/PDF paper version we show much less code but for the html vignette we could show more via code-folding.

eahowerton commented 3 months ago

This seems like a good approach to me. I've opted to keep

  1. all the code related to computing ensembles,
  2. an example of plotting with hubVis, and
  3. an example of transforming data using as_model_out_tbl.

The code that we use to generate the tables felt more distracting than helpful, as it is only filtering/formatting. I've removed this. But feel free to object.