ggplot - Centralize facet_grid title and appear only once

Question

I've created a graph in ggplot with two variables inside facet_grid.

I want the title of each facet to repeat only once and in the center of the facet.

For example, the zeros and ones in the first raw (upper facet), will appear only once and in the middle.

In my original plot, the number of plots per facet is not equal. Thus, patching together two plots using patchwork/ cowplot / ggpubr doesn't work very well.
I prefer a solution/hack only using ggplot.

Sample data:

df <- head(mtcars, 5)

Example plot:

df %>% 
  ggplot(aes(gear, disp)) + 
  geom_bar(stat = "identity") + 
  facet_grid(~am + carb,
             space = "free_x", 
             scales = "free_x") +
  ggplot2::theme(
    panel.spacing.x = unit(0,"cm"), 
    axis.ticks.length=unit(.25, "cm"), 
    strip.placement = "outside",
    legend.position = "top",
    legend.justification = "center",
    legend.direction = "horizontal",
    legend.key.size = ggplot2::unit(1.5, "lines"),
    # switch off the rectangle around symbols
    legend.key = ggplot2::element_blank(),
    legend.key.width = grid::unit(2, "lines"),
    # # facet titles
    strip.background = ggplot2::element_rect(
      colour = "black",
      fill = "white"),
    panel.background = ggplot2::element_rect(
      colour = "white",
      fill = "white"), 
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank())

Edit - new data

I created a sample data which resembles my actual data more accurately.

structure(list(par = c("Par1", "Par1", "Par1", "Par1", "Par1", 
"Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", 
"Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", 
"Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par2", "Par2", 
"Par2"), channel_1 = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 11L, 11L, 11L, 11L, 
11L, 11L, 11L, 11L, 11L, 1L, 1L, 1L), .Label = c("Center", "Left \nFrontal", 
"Left \nFrontal Central", "Left \nCentral Parietal", "Left \nParietal Ooccipital", 
"Left", "Right \nFrontal", "Right \nFrontal Central", "Right \nCentral Parietal", 
"Right \nParietal Ooccipital", "Right"), class = "factor"), freq = structure(c(1L, 
1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 
3L, 1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L), .Label = c("Alpha", 
"Beta", "Gamma"), class = "factor"), group = c("a", "b", "c", 
"a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", 
"b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", 
"c"), m = c(0.488630500442935, 0.548666228768508, 0.0441536349332613, 
0.304475866391531, 0.330039488441422, 0.0980622573307064, 0.0963996979198171, 
0.301679466108907, 0.240618782227119, 0.35779695722622, 0.156116647839907, 
0.0274546218676152, 0.0752501569920047, 0.289342864254614, 0.770518960576786, 
0.548130676907356, 0.180158614358946, 0.238520826021687, 0.406326198917495, 
0.159739769132509, 0.140739952534666, 0.295427640977557, 0.106130817023844, 
0.214006898241167, 0.31081727835652, 0.366982521446529, 0.264432086988446, 
0.0761271112139142, 0.0811642772125171, 0.0700455890939194), 
    se = c(0.00919040825504951, 0.00664655073810519, 0.0095517721611042, 
    0.00657090455386036, 0.00451135146762504, 0.0188625074573698, 
    0.00875378313351897, 0.000569521129673224, 0.00691447732630984, 
    0.000241814142091401, 0.0124584589176995, 0.00366855139256551, 
    0.0072981677277562, 0.0160663614099261, 0.00359337442316408, 
    0.00919725279757502, 0.040856967817406, 0.00240910563984416, 
    0.0152236046767608, 0.00765487375180611, 0.00354140237391633, 
    0.00145468584619171, 0.0185141245423404, 0.000833307847848054, 
    0.0038193622895167, 0.0206130436440409, 0.0066911922721337, 
    7.3079999953491e-05, 0.0246233416039572, 0.00328150956514463
    )), row.names = c(NA, -30L), class = c("tbl_df", "tbl", "data.frame"
))

Plot:

df %>%
  ggplot(aes(channel_1, m, 
             group = group,
             fill = group, 
             color = group)) +
  facet_grid(~par + freq,
             space="free_x", 
             scales="free_x") +
  geom_errorbar(
    aes(min = m - se, ymax = m + se, alpha = 0.01), 
    width = 0.2, size = 2, color = "black", 
    position = position_dodge(width = 0.6)) +
  geom_bar(stat = "identity",
           position = position_dodge(width = 0.6),
           # color = "black", 
           # fill = "white",
           width = 0.6, 
           size = 2, aes(alpha = 0.01))  + 
  scale_shape_manual(values = c(1, 8, 5)) + 
  labs(
    color = "",
    fill = "", 
    shape = "") +
  guides(
    color = FALSE,
    shape = FALSE) +
  scale_alpha(guide = "none")

Should have looked earlier for duplicates haha ! But I learned some new things so wasn't too bad :) [Nested facets in ggplot2 spanning groups](https://stackoverflow.com/questions/40316169/nested-facets-in-ggplot2-spanning-groups) — tjebo, Mar 24 '20 at 13:36

tjebo · Accepted Answer · 2020-03-24T13:41:19.493

The quickest hack: fake the facet with a plot and combine. This requires some hacking, but it's probably still less hack than messing with the grobs:

make joint variable for facet plot.
make fake facet and combine with package such as patchwork. reduce margins of plots to minus, so that there is really no margin.
make the relative height ratio kind of ridiculously high, so the second plot disappears and only facet strip remains.

library(patchwork)
library(tidyverse)

df <- head(mtcars,5)
df <- df %>% mutate(am_carb = factor(paste(am,carb,sep = '_'), 
                      labels = c( ' 1','2','1','4')))
##note!! the blank space in ' 1' label is on purpose!!! this is to make those labels unique, otherwise it would consider both '1' the same category!!

p1 <-
  df %>% 
  ggplot(aes(gear, disp)) + 
  geom_bar(stat = "identity") + 
  facet_grid(~am_carb, scales = "free_x") +
theme(panel.spacing.x = unit(0,"cm"),
      plot.margin = margin(t = -2),
      strip.background = element_rect(colour = "black",fill = "white"),
      panel.background = element_rect(colour = "white", fill = "white"), 
      panel.grid.major = element_blank(),
      panel.grid.minor = element_blank())

p2 <-
  df %>% 
  ggplot(aes(gear, disp)) + 
  geom_blank() + 
  facet_grid(~ am, scales = "free_x") +
  theme(panel.spacing.x = unit(0,"cm"),
        axis.text = element_blank(),
        axis.ticks = element_blank(),
        axis.title = element_blank(),
        plot.margin = margin(b = -2),
        strip.background = element_rect(colour = "black",fill = "white"),
        panel.background = element_rect(colour = "white", fill = "white"), 
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank())


p2/p1 + plot_layout(heights = c(0.1,100) )

^{Created on 2020-03-24 by the reprex package (v0.3.0)}

update with new data - some more complex facets. Indeed, patchwork is difficult here. Easier to combine the fake facets with cowplot, after transforming the fake facets to grid object and changing the widths. All within cowplot.

mydat <- structure(list(par = c("Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par1", "Par2", "Par2", "Par2"), channel_1 = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 1L, 1L, 1L), .Label = c("Center", "Left \nFrontal", "Left \nFrontal Central", "Left \nCentral Parietal", "Left \nParietal Ooccipital", "Left", "Right \nFrontal", "Right \nFrontal Central", "Right \nCentral Parietal", "Right \nParietal Ooccipital", "Right"), class = "factor"), freq = structure(c(1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L), .Label = c("Alpha", "Beta", "Gamma"), class = "factor"), group = c("a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c"), m = c(0.488630500442935, 0.548666228768508, 0.0441536349332613, 0.304475866391531, 0.330039488441422, 0.0980622573307064, 0.0963996979198171, 0.301679466108907, 0.240618782227119, 0.35779695722622, 0.156116647839907, 0.0274546218676152, 0.0752501569920047, 0.289342864254614, 0.770518960576786, 0.548130676907356, 0.180158614358946, 0.238520826021687, 0.406326198917495, 0.159739769132509, 0.140739952534666, 0.295427640977557, 0.106130817023844, 0.214006898241167, 0.31081727835652, 0.366982521446529, 0.264432086988446, 0.0761271112139142, 0.0811642772125171, 0.0700455890939194), se = c(0.00919040825504951, 0.00664655073810519, 0.0095517721611042, 0.00657090455386036, 0.00451135146762504, 0.0188625074573698, 0.00875378313351897, 0.000569521129673224, 0.00691447732630984, 0.000241814142091401, 0.0124584589176995, 0.00366855139256551, 0.0072981677277562, 0.0160663614099261, 0.00359337442316408, 0.00919725279757502, 0.040856967817406, 0.00240910563984416, 0.0152236046767608, 0.00765487375180611, 0.00354140237391633, 0.00145468584619171, 0.0185141245423404, 0.000833307847848054, 0.0038193622895167, 0.0206130436440409, 0.0066911922721337, 7.3079999953491e-05, 0.0246233416039572, 0.00328150956514463)), row.names = c(NA, -30L), class = c("tbl_df", "tbl", "data.frame"))

library(tidyverse)
library(cowplot)
#> 
#> ********************************************************
#> Note: As of version 1.0.0, cowplot does not change the
#>   default ggplot2 theme anymore. To recover the previous
#>   behavior, execute:
#>   theme_set(theme_cowplot())
#> ********************************************************

mydat <- mydat %>% mutate(par_freq = factor(paste(par,freq,sep = '_'), labels = c('Alpha', 'Beta', 'Gamma', 'Gamma ' )))

p1 <-
  mydat %>% 
    ggplot(aes(channel_1, m, group = group, fill = group, color = group)) +
  geom_bar(stat = "identity") + 
  facet_grid( ~ par_freq, scales = "free_x", space="free_x") +
  theme(panel.spacing.x = unit(0,"cm"),
        plot.margin = margin(t = -2),
        strip.background = element_rect(colour = "black",fill = "white"),
        panel.background = element_rect(colour = "white", fill = "white"), 
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        legend.position = 'none')

p2 <-
  mydat %>%
    ggplot(aes(channel_1, m, group = group, fill = group, color = group)) +
    geom_blank() + 
    facet_grid(~ par) +
    theme(panel.spacing.x = unit(0,"cm"),
          axis.text = element_blank(),
          axis.ticks = element_blank(),
          axis.title = element_blank(),
          plot.margin = margin(b = -2),
          strip.background = element_rect(colour = "black",fill = "white"),
          panel.background = element_rect(colour = "white", fill = "white"), 
          panel.grid.major = element_blank(),
          panel.grid.minor = element_blank())


gt <-  cowplot::as_gtable(p2)
gt$widths[5] <- 8*gt$widths[7]
cowplot::plot_grid(gt, p1, align = "v", axis = 'l',nrow = 2, rel_heights = c(5, 100))
# you need to play around with the values unfortunately.

^{Created on 2020-03-24 by the reprex package (v0.3.0)}

Some additional thoughts

I was thinking that one cannot get around a hack like that - because the gtable_layout of the original plot (with two facet variables) shows the entire facet strip is one grob! THis answer proved me wrong - the grob contains a nested table for both strips!. But there is an easier solution thanks to the ggnomics package - see my second answer

p_demo <-   ggplot(mydat, aes(channel_1, m)) + 
  geom_bar(stat = "identity") + 
  facet_grid(~par +freq , space = "free_x",  scales = "free_x") +
  theme(panel.spacing.x = unit(0,"cm"))

gt <-  cowplot::as_gtable(p_demo)
gtable::gtable_show_layout(gt)

^{Created on 2020-03-24 by the reprex package (v0.3.0)}

`patchwork` is a fantastic package. However, my problem is that my original two graphs are not in the same size. Thus, if I don't use facet_grid together the ratios looks really bad. — DJV, Mar 24 '20 at 07:29
God, that's a nightmare ;). I hope it won't be difficult to reproduce this again!. What I actually started to look at, with no luck, is to just draw rectangles over the facet titles (https://stackoverflow.com/questions/60827811/ggplot-cover-facet-grid-title-with-rectangle) — DJV, Mar 24 '20 at 12:58

score 1 · Answer 2 · answered Mar 24 '20 at 13:31

Sorry to add a second answer, but I believe it's different enough to merit a separate answer. I should have thought of the ggnomics package earlier, which makes this task super simple!

 #devtools::install_github("teunbrand/ggnomics")
  library(ggnomics)
#> Loading required package: ggplot2
  library(tidyverse)

  mydat<- head(mtcars, 5)
  mydat %>% 
    ggplot(aes(gear, disp)) + 
    geom_bar(stat = "identity") + 
    facet_nested(~am + carb) +
    theme(panel.spacing.x = unit(0,"cm"), 
          axis.ticks.length=unit(.25, "cm"), 
          strip.placement = "inside",
          strip.background = element_rect( colour = "black", fill = "white"),
          panel.background = element_rect( colour = "black", fill = "white"))

^{Created on 2020-03-24 by the reprex package (v0.3.0)}

And btw, after getting an error, it seems like there are clashes between `ggnomics` and the new ggplot (https://github.com/teunbrand/ggnomics/issues/37). Thus, the author says to use the package `ggh4x` (https://github.com/teunbrand/ggh4x), if you want to use `facet_nested`. And thank you again! — DJV, Mar 24 '20 at 14:17

ggplot - Centralize facet_grid title and appear only once

Edit - new data

2 Answers2

Linked