How to put axes behind the graph?

Question

I created a graph using geom_line and geom_point via ggplot. I want my axes to meet at (0,0) and I want my lines and data points to be in front of the axes instead of behind as shown:

I've tried:

1. coord_cartesian(clip = 'off')
2. putting geom_line and geom_point at the end
3. creating a base graph then add geom_line and geom_point
4. playing around with the functions of coord_cartesian
5. manually setting xlim =c(-0.1, 25) and ylim=c(-0.1, 1500)

data7 is as follows:

   Treatment Days N        mean         sd          se
1          1    0 7 204.7000000  41.579963  15.7157488
2          1    2 7 255.0571429  41.116617  15.5406205
3          1    5 7 290.6000000  49.506498  18.7116974
4          1    8 7 330.8142857  49.044144  18.5369442
5          1   12 7 407.5142857  95.584194  36.1274294
6          1   15 7 540.8571429 164.299390  62.0993323
7          1   19 7 737.5285714 308.786359 116.7102736
8          1   21 7 978.4571429 502.506726 189.9296898
9          2    0 7 205.7428571  46.902482  17.7274721
10         2    2 7 227.5571429  47.099889  17.8020846
11         2    5 7 232.4857143  59.642922  22.5429054
12         2    8 7 247.9857143  66.478529  25.1265220
13         2   12 7 272.0428571  79.173162  29.9246423
14         2   15 7 289.1142857  82.847016  31.3132288
15         2   19 7 312.3857143 105.648591  39.9314140
16         2   21 7 334.7142857 121.569341  45.9488920
17         3    0 7 212.2285714  47.549263  17.9719320
18         3    2 7 235.4142857  52.689671  19.9148237
19         3    5 7 177.0714286  54.895225  20.7484447
20         3    8 7 205.2571429  72.611451  27.4445489
21         3   12 7 247.8142857 119.369558  45.1174522
22         3   15 7 280.4285714 140.825847  53.2271669
23         3   19 7 366.9142857 210.573799  79.5894149
24         3   21 7 451.0428571 289.240793 109.3227438
25         4    0 7 211.6857143  24.329161   9.1955587
26         4    2 7 227.8428571  28.762525  10.8712127
27         4    5 7 205.9428571  49.148919  18.5765451
28         4    8 7 153.1142857  25.189246   9.5206399
29         4   12 7 128.2571429  43.145910  16.3076210
30         4   15 7 104.1714286  45.161662  17.0695038
31         4   19 7  85.4714286  51.169708  19.3403318
32         4   21 7  66.9000000  52.724567  19.9280133
33         5    0 7 216.7857143  39.957829  15.1026398
34         5    2 7 212.2000000  27.037135  10.2190765
35         5    5 7 115.5000000  37.094070  14.0202405
36         5    8 7  46.1000000  34.925492  13.2005952
37         5   12 7  29.3142857  24.761222   9.3588621
38         5   15 6  10.0666667  13.441974   5.4876629
39         5   19 6   6.4000000  11.692733   4.7735382
40         5   21 6   5.3666667  12.662017   5.1692467
41         6    0 7 206.6857143  40.359155  15.2543269
42         6    2 7 197.0428571  40.608327  15.3485048
43         6    5 7 106.2142857  58.279654  22.0276388
44         6    8 7  46.0571429  62.373014  23.5747833
45         6   12 7  31.7571429  49.977457  18.8897031
46         6   15 7  28.1142857  45.437995  17.1739480
47         6   19 7  26.2857143  38.414946  14.5194849
48         6   21 7  32.7428571  53.203003  20.1088450
49         7    0 7 193.2000000  37.300447  14.0982437
50         7    2 7 133.2428571  26.462606  10.0019250
51         7    5 7   3.8142857   7.445900   2.8142857
52         7    8 7   0.7142857   1.496026   0.5654449
53         7   12 7   0.0000000   0.000000   0.0000000
54         7   15 7   0.0000000   0.000000   0.0000000
55         7   19 7   0.0000000   0.000000   0.0000000
56         7   21 7   0.0000000   0.000000   0.0000000

My code is as follows:

ggplot(data7, aes(Days, mean, color=Treatment)) +
  geom_line() +
  geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=0.5, size= 0.25) +
  geom_point(size=2.5) +
  scale_colour_hue(limits = c("1", "2", "3", "4", "5", "6", "7")) +
  scale_x_continuous(expand = c(0, 0), limits = c(0, NA), breaks = scales::pretty_breaks(n = 10)) +
  scale_y_continuous(expand = c(0, 0), limits = c(0, NA), breaks = scales::pretty_breaks(n = 8)) +
  theme_classic() +
  theme(axis.text = element_text(color = "#000000"), plot.title = element_text(hjust = 0.5)) +
  coord_cartesian(clip = 'off')

Levine

Can you either post enough sample data to reproduce this, or shift to a base-R dataset? – Lantz

@Lantz I have edited the post with the data! Thanks for helping me out! – Levine

It might be just me, but ... my y-axis is not at x=0, it's a little left of it. – Lantz

@Lantz Hmm, what about your x-axis? I ideally would like the axes to cross exactly at 0,0 and having data start from the axes instead of having them float around if that makes sense, haha. – Levine

Fortunately (for me), my answer below forces the 0,0 intersection :-), hope it works (or at least helps). – Lantz

Answer 1

Here's one approach that omits the axis lines/ticks and then explicitly layers them below the rest of the plot layers. Because the new lines/ticks are drawn as literal objects, they will then ignore any other theming you may later apply. With control comes responsibility ...

This method has the side-effect of a "simple" axis tick, just the + symbol, which shows as a cross-line at each point. This is in contrast to the standard way (typically just pointing outwards). I'm guessing that something more robust could be devised, but I thought "simple" up-front could be adapted in other ways.

Taking the literal code of your ggplot(...) + ... and storing as gg, no changes. First we'll extract the tick marks. If you are confident enough (or not OCD-enough) to determine the tick locations yourself, then feel free to hard-code it. This method (of using ggplot_build then extracting the ...$x$breaks) has the advantage of matching the tick and label locations, especially if they might change with different/updated data.

ticks <- with(ggplot_build(gg)$layout$panel_params[[1]],
              na.omit(rbind(
                data.frame(x = x$breaks, y = 0),
                data.frame(x = 0, y = y$breaks)
              )))
head(ticks,3); tail(ticks,3)
#   x y
# 1 0 0
# 2 2 0
# 3 4 0
#    x    y
# 16 0  600
# 17 0  800
# 18 0 1000

From here, I'll take a cue from https://mcmap.net/q/383875/-insert-layer-underneath-existing-layers-in-ggplot2-object and prepend some layers below all of the others. (This is where I identify the + symbol for axis ticks, using shape=3.)

gg$layers <- c(
  geom_hline(aes(yintercept = 0)),
  geom_vline(aes(xintercept = 0)),
  geom_point(data = ticks, aes(x, y), shape = 3, inherit.aes = FALSE),
  gg$layers)

Now we just plot the previously-generated gg, adding a cue to omit the theme axis lines/ticks.

gg + theme(axis.line = element_blank(), axis.ticks = element_blank())

---

Data, including converting Treatment to character (to avoid continuous/discrete warnings from scale_colour_hue):

data7 <- read.table(header=TRUE, text = "
   Treatment Days N        mean         sd          se
1          1    0 7 204.7000000  41.579963  15.7157488
2          1    2 7 255.0571429  41.116617  15.5406205
3          1    5 7 290.6000000  49.506498  18.7116974
4          1    8 7 330.8142857  49.044144  18.5369442
5          1   12 7 407.5142857  95.584194  36.1274294
6          1   15 7 540.8571429 164.299390  62.0993323
7          1   19 7 737.5285714 308.786359 116.7102736
8          1   21 7 978.4571429 502.506726 189.9296898
9          2    0 7 205.7428571  46.902482  17.7274721
10         2    2 7 227.5571429  47.099889  17.8020846
11         2    5 7 232.4857143  59.642922  22.5429054
12         2    8 7 247.9857143  66.478529  25.1265220
13         2   12 7 272.0428571  79.173162  29.9246423
14         2   15 7 289.1142857  82.847016  31.3132288
15         2   19 7 312.3857143 105.648591  39.9314140
16         2   21 7 334.7142857 121.569341  45.9488920
17         3    0 7 212.2285714  47.549263  17.9719320
18         3    2 7 235.4142857  52.689671  19.9148237
19         3    5 7 177.0714286  54.895225  20.7484447
20         3    8 7 205.2571429  72.611451  27.4445489
21         3   12 7 247.8142857 119.369558  45.1174522
22         3   15 7 280.4285714 140.825847  53.2271669
23         3   19 7 366.9142857 210.573799  79.5894149
24         3   21 7 451.0428571 289.240793 109.3227438
25         4    0 7 211.6857143  24.329161   9.1955587
26         4    2 7 227.8428571  28.762525  10.8712127
27         4    5 7 205.9428571  49.148919  18.5765451
28         4    8 7 153.1142857  25.189246   9.5206399
29         4   12 7 128.2571429  43.145910  16.3076210
30         4   15 7 104.1714286  45.161662  17.0695038
31         4   19 7  85.4714286  51.169708  19.3403318
32         4   21 7  66.9000000  52.724567  19.9280133
33         5    0 7 216.7857143  39.957829  15.1026398
34         5    2 7 212.2000000  27.037135  10.2190765
35         5    5 7 115.5000000  37.094070  14.0202405
36         5    8 7  46.1000000  34.925492  13.2005952
37         5   12 7  29.3142857  24.761222   9.3588621
38         5   15 6  10.0666667  13.441974   5.4876629
39         5   19 6   6.4000000  11.692733   4.7735382
40         5   21 6   5.3666667  12.662017   5.1692467
41         6    0 7 206.6857143  40.359155  15.2543269
42         6    2 7 197.0428571  40.608327  15.3485048
43         6    5 7 106.2142857  58.279654  22.0276388
44         6    8 7  46.0571429  62.373014  23.5747833
45         6   12 7  31.7571429  49.977457  18.8897031
46         6   15 7  28.1142857  45.437995  17.1739480
47         6   19 7  26.2857143  38.414946  14.5194849
48         6   21 7  32.7428571  53.203003  20.1088450
49         7    0 7 193.2000000  37.300447  14.0982437
50         7    2 7 133.2428571  26.462606  10.0019250
51         7    5 7   3.8142857   7.445900   2.8142857
52         7    8 7   0.7142857   1.496026   0.5654449
53         7   12 7   0.0000000   0.000000   0.0000000
54         7   15 7   0.0000000   0.000000   0.0000000
55         7   19 7   0.0000000   0.000000   0.0000000
56         7   21 7   0.0000000   0.000000   0.0000000")
data7$Treatment <- as.character(data7$Treatment)

Lantz

Answer 2

A fairly straightforward way to do this is just to move the panel in front of the axes once the plot elements are created (i.e. as a grobTree). The grobTree contains a layout data frame which allows you to move plot elements forwards or backwards by adjusting their z component.

If you store your plot as p, then the code would be:

ggp <- ggplot_gtable(ggplot_build(p))
ggp$layout$z[which(ggp$layout$name == "panel")] <- max(ggp$layout$z) + 1
grid::grid.draw(ggp)

---

Plot code:

This is just the original plot except I have added a vline at 0 and an hline at 0 in case bringing the panel forwards clips your axis lines).

p <- ggplot(data7, aes(Days, mean, color=Treatment)) +
  geom_hline(aes(yintercept = 0)) +
  geom_vline(aes(xintercept = 0)) +
  geom_line() +
  geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=0.5, size= 0.25) +
  geom_point(size=2.5) +
  scale_colour_hue(limits = c("1", "2", "3", "4", "5", "6", "7")) +
  scale_x_continuous(expand = c(0, 0), limits = c(0, NA), breaks = scales::pretty_breaks(n = 10)) +
  scale_y_continuous(expand = c(0, 0), limits = c(0, NA), breaks = scales::pretty_breaks(n = 8)) +
  theme_classic() +
  theme(axis.text = element_text(color = "#000000"), plot.title = element_text(hjust = 0.5)) +
  coord_cartesian(clip = 'off')

Nyeman