Final preflight checks

[gavinsimpson]

Issue for anything we want to do before we resubmit, but which isn't a trivial edit to fix a typo. This is an incomplete list based on what I have noticed from a scan through yesterday; please add more things here if you notice them so we don't forget.

• [x] Rebuild from scratch on several systems and all OK? #44
  • [x] @eric-pedersen (Windows?)
  • [x] @dill (Mac)
  • [x] @gavinsimpson (Linux)
• [x] Tidy code blocks Some of the code blocks could use the full line more efficiently and be tidied/aligned a little better. fddd808
• [x] Add my (@gavinsimpson) Biology affiliation
• [x] Move legend on Figure 18 panel b to the top of the plot - addressed in d98c28f
• [x] Move Legend for Figure 17 to top of plot to match Figure 15 - addressed in d98c28f
• [x] Alter Table 2? - addressed in d98c28f Currently we're just dumping the output from k.check() into a table. This would look nicer if we modified the column headers to be k' (i.e. proper prime) and "EDF" to match the test. Add custom row labels to remove the R/mgcv-specific encoding. For example we use f for smooth function in equations so this should not really be literally using s(....). Something like "f(day) C. sphaericus" for the day by taxon smooths, "f(taxon)" and "f(taxon, year)" for the taxon and taxon by year random effects, with appropriate LaTeX to match the way the equations are written.
• [x] Tweak point size on some figures (slightly smaller) - addressed in d98c28f
  • [x] Figure 14
  • [x] Figure 16
• [x] In Table 3, correct "taxon" -> "Taxon" as this is a heading - addressed in d98c28f
• [x] Yellow line in Figure 3b, basis example, is very difficult to read on the white background when printed. Truncate the discrete palette so it doesn't extending into the yellows? - addressed in d98c28f
• [x] Determine why table 5 no longer seems to be calculating smooths appropriately #50
• [x] Draft response to reviewers
• [x] Have all of authors sign off on response.
  • [x] @eric-pedersen
  • [x] @noamross
  • [x] @dill
  • [x] @gavinsimpson

[eric-pedersen]

Thanks for putting this together, Gavin! I just updated the to-do list with going through the response to reviewers. I started working on that yesterday as a Google Doc, so hopefully will have it together soon.

[eric-pedersen]

I've altered Table 2 as suggested, but I couldn't figure out how to get a prime symbol to render in the table; it kept rendering as a backwards single quotation mark. I think for now it's okay to leave it as is, as that is how the term appears in mgcv, and it shouldn't cause too much confusion in the table itself.

[eric-pedersen]

I've gone through and I think implemented all of these changes. @gavinsimpson, want to check to make sure you're good with all the tweaks? Especially table 2; I wasn't able to get the k' working properly in the table, but maybe someone with better Rmarkdown-fu can figure it out.

I decided to go with a fully different colour scheme for figure 2 in commit 3176e64, as the discrete viridis scale is too difficult to distinguish without the yellow end. I found a good discrete scale, with clear visual separation of categories, in ggthemes, and have implemented that. What do you think of this new scheme?

Also, I've gone through the code again to see any places that could be cleaned up, but this could use a second or third pair of eyes.

[gavinsimpson]

That looks great @eric-pedersen

• could just do with aligning the bottom row axes with the upper panel's
• if this is from the paper render it doesn't seem to be picking up the theme we are using

I can take a look at the first one when I do a build a read through this PM. I'm just trying to fix a little bug in gratia which is leading to inconsistent results and is stopping me submitting to CRAN.

[eric-pedersen]

I've fixed this now in the most recent commit, and updated the figure above to show the new alignment (also: plot_grid is such a neat function!). Does the updated figure look good to you?

[gavinsimpson]

I've altered Table 2 as suggested, but I couldn't figure out how to get a prime symbol to render in the table; it kept rendering as a backwards single quotation mark. I think for now it's okay to leave it as is, as that is how the term appears in mgcv, and it shouldn't cause too much confusion in the table itself.

I tried to take a look and it seems that there is some incompatibility with how we are rendering the tables, injecting latex into the columns names, and our rendering format. We need to have escape = FALSE for the table to get the latex rendering but if we do that, now matter how I fiddled with things I couldn't get texi2dvi to complete the build of the PDF without erroring out. So agreed; leave this as is now.