Multiple ggplots of different sizes

Question

It's relatively simple using grid.arrange in the gridExtra package to arrange multiple plots in a matrix, but how can you arrange plots (the ones I'm working on are from ggplot2) when some plots are intended to be larger than others? In base, I can use layout() such as in the example below:

 nf <- layout(matrix(c(1,1,1,2,3,1,1,1,4,5,6,7,8,9,9), byrow=TRUE, nrow=3))
 layout.show(nf)

what is the equivalent for ggplot plots?

enter image description here

Some plots for inclusion

library(ggplot2)
p1 <- qplot(x=wt,y=mpg,geom="point",main="Scatterplot of wt vs. mpg", data=mtcars)
p2 <- qplot(x=wt,y=disp,geom="point",main="Scatterplot of wt vs disp", data=mtcars)
p3 <- qplot(wt,data=mtcars)
p4 <- qplot(wt,mpg,data=mtcars,geom="boxplot")
p5 <- qplot(wt,data=mtcars)
p6 <- qplot(mpg,data=mtcars)
p7 <- qplot(disp,data=mtcars)
p8 <- qplot(disp, y=..density.., geom="density", data=mtcars)
p9 <- qplot(mpg, y=..density.., geom="density", data=mtcars)

This [SO question](http://stackoverflow.com/questions/9490482/combined-plot-of-ggplot2-not-in-a-single-plot-using-par-or-layout-functio) could help. — Didzis Elferts, Aug 25 '13 at 09:24

Roland · Answer 1 · 2013-08-26T18:22:53.530

13

You can use nested arrangeGrob calls like this example:

library(ggplot2)
library(gridExtra)

p <- ggplot(data.frame(x=1, y=1), aes(x,y)) + geom_point()

grid.arrange(
  arrangeGrob(
    p, 
    arrangeGrob(p, p, nrow=2),
    ncol=2 ,widths=c(2,1)),
  arrangeGrob(p, p ,p ,ncol=3, widths=rep(1,3)),
  nrow=2)

Edit:

gl <- lapply(1:9, function(ii) grobTree(rectGrob(),textGrob(ii)))

grid.arrange(
  arrangeGrob(gl[[1]],
              do.call(arrangeGrob, c(gl[2:5], ncol=2)),
              nrow=1,
              widths=3:2),
  do.call(arrangeGrob, c(gl[6:9], nrow=1, list(widths=c(1,1,1,2)))),
nrow=2, heights=c(2,1))

enter image description here

edited Aug 26 '13 at 18:22

answered Aug 25 '13 at 09:29

Roland

127,288
10
191
288

you don't need an anonymous function with `do.call`, there's already `...` in the definition of `arrangeGrob` – baptiste Aug 25 '13 at 13:57
@baptiste I know and use that with `nrow`, but got an error when trying to pass `widths`. – Roland Aug 25 '13 at 14:01
1

it's because of the way `c()` works with lists. One needs to use `do.call(arrangeGrob, c(gl[6:9], nrow=1, list(widths=c(1,1,1,2))))` – baptiste Aug 25 '13 at 21:13

score 11 · Answer 2 · answered Aug 25 '13 at 12:14

11

An alternative with gtable

library(gtable)

gl <- lapply(1:9, function(ii) grobTree(textGrob(ii), rectGrob()))
# gl <- lapply(1:9, function(ii) ggplotGrob(qplot(1,1) + ggtitle(ii)))

gt <- gtable(widths=unit(rep(1,5), "null"),
             heights=unit(rep(1,3), "null"))

gtable_add_grobs <- gtable_add_grob # alias

gt <- gtable_add_grobs(gt, gl, 
                       l=c(1,4,5,4,5,1,2,3,4),
                       r=c(3,4,5,4,5,1,2,3,5),
                       t=c(1,1,1,2,2,3,3,3,3),
                       b=c(2,1,1,2,2,3,3,3,3))
grid.newpage()
grid.draw(gt)

enter image description here

answered Aug 25 '13 at 12:14

baptiste

75,767
19
198
294

This seems a popular method, but I can't work out how to add ggplot-like grobs to the plot. (In Roland's example, it's easy.) Could you explain how to add plots in? (I've included some sample plots in OP.) – Hugh Aug 26 '13 at 23:54
1

`ggplotGrob` converts a ggplot to a grob, there's a commented line using it in my example – baptiste Aug 27 '13 at 00:01

baptiste · Answer 3 · 2016-06-06T22:00:20.417

You can use the same matrix interface as layout with grid.arrange,

library(gridExtra)
library(grid)
gl <- lapply(1:9, function(ii) grobTree(rectGrob(), textGrob(ii)))

grid.arrange(grobs = gl, layout_matrix = rbind(c(1,1,1,2,3),
                                               c(1,1,1,4,5),
                                               c(6,7,8,9,9)))

enter image description here

and the same works for ggplots; note that NA can be used to indicate blank cells. The result is a gtable, compatible with ggsave().

gl <- replicate(9, ggplot(), FALSE)
grid.arrange(grobs = gl, layout_matrix = rbind(c(1,1,1,2,3),
                                               c(1,1,1,4,5),
                                               c(6,7,8,NA,9)))

score 8 · Accepted Answer · edited May 23 '17 at 12:17

I appreciate all the other answers, but Didzis Elferts's comment on the OP connected to the answer that I found easiest to implement.

library(ggplot2)
p1 <- qplot(x=wt,y=mpg,geom="point",main="Scatterplot of wt vs. mpg", data=mtcars)
p2 <- qplot(x=wt,y=disp,geom="point",main="Scatterplot of wt vs disp", data=mtcars)
p3 <- qplot(wt,data=mtcars)
p4 <- qplot(wt,mpg,data=mtcars,geom="boxplot")
p5 <- qplot(wt,data=mtcars)
p6 <- qplot(mpg,data=mtcars)
p7 <- qplot(disp,data=mtcars)
p8 <- qplot(disp, y=..density.., geom="density", data=mtcars)
p9 <- qplot(mpg, y=..density.., geom="density", data=mtcars)

vplayout <- function(x, y) viewport(layout.pos.row = x, layout.pos.col = y)

grid.newpage()
pushViewport(viewport(layout = grid.layout(3, 5))) # 3 rows, 5 columns
print(p1, vp = vplayout(1:2, 1:3))  # the big plot covers rows 1:2 and cols 1:3
print(p2, vp = vplayout(1, 4))
print(p3, vp = vplayout(1, 5))
print(p4, vp = vplayout(2, 4))
print(p5, vp = vplayout(2, 5))
print(p6, vp = vplayout(3, 1))
print(p7, vp = vplayout(3, 2))
print(p8, vp = vplayout(3, 3))
print(p9, vp = vplayout(3, 4:5))

This is great! Any way to add a single title at the top of the superplot? — Pertinax, May 13 '18 at 14:23

Gregor Thomas · Answer 5 · 2016-06-06T16:04:52.120

8

I like the interface provided by the lay_out function (formerly in the wq package) . It takes arguments of the form list(plot, row(s), column(s)). For your example:

lay_out(list(p1, 1:2, 1:3),
       list(p2, 1, 4),
       list(p3, 1, 5),
       list(p4, 2, 4),
       list(p5, 2, 5),
       list(p6, 3, 1),
       list(p7, 3, 2),
       list(p8, 3, 3),
       list(p9, 3, 4:5))

Which yields:

enter image description here

lay_out = function(...) {    
    x <- list(...)
    n <- max(sapply(x, function(x) max(x[[2]])))
    p <- max(sapply(x, function(x) max(x[[3]])))
    grid::pushViewport(grid::viewport(layout = grid::grid.layout(n, p)))    

    for (i in seq_len(length(x))) {
        print(x[[i]][[1]], vp = grid::viewport(layout.pos.row = x[[i]][[2]], 
            layout.pos.col = x[[i]][[3]]))
    }
}

(Code sourced from a prior version of the wq package, from the commit history on the unofficial Github CRAN mirror.)

edited Jun 06 '16 at 16:04

answered Jul 25 '14 at 18:38

Gregor Thomas

136,190
20
167
294

this is the simplest solution but it appears that the layOut function no longer exists. Are there any other straightforward options that you are aware of? – Meli Jun 06 '16 at 15:30
Thank you. Is there a way to make this work for combining ggplot and non ggplot graphics? – Meli Jun 10 '16 at 18:51
It should work for any `grid` graphics, which mostly means `ggplot` and `lattice`. If you want to use `base` graphics you could maybe make it work with the `gridBase` package. – Gregor Thomas Jun 10 '16 at 19:48

Multiple ggplots of different sizes

5 Answers5

Linked

Related