Bug in facetted ggplot with coord.fixed()

I think this is a bug, others disagree!

Explanation

The issue is that I get a lot of unwanted white space above and below facetted plots but not with plots without the facetting.

Show code

valN <- 2000
valCorr <- .7
matCov <- matrix(c(1, valCorr, valCorr, 1), byrow = TRUE, ncol = 2)
vecMeans <- c(0, 0)
MASS::mvrnorm(n = valN, mu = vecMeans, Sigma = matCov) %>%
  as_tibble(.name_repair = "universal") %>%
  rename(Pre = `...1`,
         Post = `...2`) %>%
  mutate(ID = row_number(),
         group = ID %% 2) -> tibDat

Simple plot: no unnecessary white space.

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed()

Add facets: no unwanted white space.

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) 

Adding coord_fixed() causes the issue

This happens after adding facetting and adding coord_fixed().

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed()

Same with theme(aspect.ratio = 1)

There was a suggestion on the web, but about a slightly different issue about scales = "free" in the facetting that using theme(aspect.ratio = 1) was better than coord_fixed(). That doesn’t change things for me.

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_wrap(facets = vars(group)) +
  geom_point(alpha = .1) +
  geom_smooth() +
  theme(aspect.ratio = 1)

Using `theme(plot.margin = …): no good

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed() +
  theme(plot.margin = margin(t = 0,
                             b = 0))

Using theme(plot.margin = margin(t = -70, b = -70)) also has no effect.

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed() +
  theme(plot.margin = margin(t = -70,
                             b = -70))

Set margin units explicitly with theme(plot.margin = unit(c(-5, 5, -5, 5), "cm")), still no joy on the vertical space though it can be seen that the right and left margins are imposed correctly.

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed() +
  theme(plot.margin = unit(c(-5, 5, -5, 5), "cm"))

Solution/bodge courtesy of stefan & axeman!

The solution is to use save the graphic to a file and then to use knitr::plot_crop() to remove the white space from the graphic and then embed the cropped file where you want it. Here’s the code:

Show code

ggplot(data = tibDat,
       aes(x = Pre, y = Post)) +
  facet_grid(cols = vars(group),
             space = "free",
             ) +
  geom_point(alpha = .1) +
  geom_smooth() +
  xlim(c(-4, 4)) +
  ylim(c(-4, 4)) +
  coord_fixed() -> tmpPlot

ggsave(filename = paste0(postDir, "tmpPlot.png"), plot = tmpPlot)
knitr::plot_crop(paste0(postDir, "tmpPlot.png")) -> tmpPng

ggsave(filename = paste0(postDir, "tmpPlot.jpg"), plot = tmpPlot)
knitr::plot_crop(paste0(postDir, "tmpPlot.jpg")) -> tmpJpg

knitr insertion, PNG

With alternative text, i.e. !![cropped_image](`r tmpPng`)

cropped_image

Without alternative text, i.e. !![](`r tmpPng`)

knitr insertion, JPEG

With alternative text, i.e. !![cropped_image](`r tmpJpg`)

Without alternative text, i.e. !![](`r tmpJpg`)

Direct insertion

For the png file, using <img src=`r tmpPng` alt=“MyPng”>

For the png file, using <img src=`r tmpJpg` alt=“MyJpg”>

Conclusion

I got the knitr::plot_crop() solution from the answers to my stackoverflow post from stefan and axeman. Thanks to them for that. I have unpacked the details about using knitr::plot_crop() from the pointers they gave to make it easier for others hitting this to make the fix. It’s largely semantic but I disagree with them and still feel that this is a bug or very much out of the grammar of graphics style of ggplot that this is the only way to fix this. To me coord_fixed() is a modification of an “inner plot” when applied after facetting and should apply to the aspect ratio of the facets not to the containing plot. Maybe I will submit it as a request for ggplot but I’m out of energy and time for that now and will just update this file and respond to the stackoverflow post.

Links

• I have put this in my Rblog as Bug in facetted ggplot with coord.fixed().
• I have also uploaded the simple Rmd reprex and its html results.
• And I have posted it here in stackoverflow.

Visit count

widget counter

Dates

• 11.vii.24 Solution/bodge added (and months in next two lines corrected!)
• 10.vii.24 Links to Rblog and raw files added, and to stackoverflow.
• 10.vii.24 First created.