How can I obtain an 'unbalanced' grid of ggplots?

94
votes

With grid.arrange I can arrange multiple ggplot figures in a grid to achieve a multi-panel figure by using something like:

library(ggplot2)
library(grid)
library(gridExtra)

generate some ggplot2 plots , then

plot5 <- grid.arrange(plot4, plot1, heights=c(3/4, 1/4), ncol=1, nrow=2)

How can I obtain an 'unbalanced' 2 col layout with one plot in the entire first col and three plots in the second col? I toyed with a 'grid-of-grids' approach by trying to use grid.arrange to plot one grid (e.g. plot5, above) against another plot, but obtained:

Error in arrangeGrob(..., as.table = as.table, clip = clip, main = main, : input must be grobs!

Update:

Thanks for the advice. I will look into viewports and grid. In the meantime, thanks to @DWin, the layOut function in the 'wq' package worked very well for the compilation figure in my Sweave document: enter image description here

Update 2:

The arrangeGrobcommand (as suggested by @baptiste) also works well, and seems very intuitive - at least it was easy to alter widths of the two columns. It also has the benefit of not requiring the `wq' package.

e.g. Here is the code from my Sweave file:

<<label=fig5plot, echo=F, results=hide>>=
plot5<-grid.arrange(plot4, arrangeGrob(plot1, plot2, plot3, ncol=1), 
                    ncol=2, widths=c(1,1.2))
@
\begin{figure}[]
    \begin{center}
<<label=fig5,fig=TRUE,echo=T, width=10,height=12>>=
<<fig5plot>>
@
\end{center}
\caption{Combined plots using the `arrangeGrob' command.}
\label{fig:five}
\end{figure}

which produces the following output: enter image description here

BTW, Anyone tell me why the '>NA' appears?

4
rggplot2r-grid
You might have to set up the viewports yourself -- grid.arrange might not be flexible enough (search stackoverflow for "[r] grid viewport") – Ben Bolker
@BenBolker Has pointed you in a fruitful direction using grid. See also Hadley's ggplot2 book, Section 8.4.2. – Ari B. Friedman
@BenBolker grid.arrange can be used with nested viewports using its companion arrangeGrob (essentially returning a gTree), as in the example I gave below. – baptiste
your final assignment to plot5 is not required as grid.arrange returns nothing (NULL). If you want to save the resulting grob use arrangeGrob again (and grid.draw to display it). – baptiste

4 Answers

73
votes

grid.arrange draws directly on the device; if you want to combine it with other grid objects you need arrangeGrob, as in 

 p = rectGrob()
 grid.arrange(p, arrangeGrob(p,p,p, heights=c(3/4, 1/4, 1/4), ncol=1),
              ncol=2)

Edit (07/2015): with v>2.0.0 you can use the layout_matrix argument,

 grid.arrange(p,p,p,p, layout_matrix = cbind(c(1,1,1), c(2,3,4)))
17
votes

I tried figuring it out with grid and thought I had it down but ended up failing (although looking now at the code in the function I cite below, I can see that I was really close ... :-)

The 'wq' package has a layOut function that will do it for you:

p1 <- qplot(mpg, wt, data=mtcars)
layOut(list(p1, 1:3, 1),   # takes three rows and the first column
        list(p1, 1, 2),    # next three are on separate rows
         list(p1, 2,2), 
          list(p1, 3,2))

enter image description here

2
votes

Another alternative is the patchwork package by Thomas Lin Pedersen.

# install.packages("devtools")
# devtools::install_github("thomasp85/patchwork")
library(patchwork)

Generate some plots.

p1 <- ggplot(mtcars) + geom_point(aes(mpg, disp)) + facet_grid(rows = vars(gear))
p2 <- ggplot(mtcars) + geom_boxplot(aes(gear, disp, group = gear))
p3 <- ggplot(mtcars) + geom_smooth(aes(disp, qsec))
p4 <- ggplot(mtcars) + geom_bar(aes(carb))

Now arrange the plots.

p1 + (p2 / p3 / p4)

enter image description here

1
votes

There is also multipanelfigure package that is worth to mention. See also this answer.

library(ggplot2)
theme_set(theme_bw())

q1 <- ggplot(mtcars) + geom_point(aes(mpg, disp))
q2 <- ggplot(mtcars) + geom_boxplot(aes(gear, disp, group = gear))
q3 <- ggplot(mtcars) + geom_smooth(aes(disp, qsec))
q4 <- ggplot(mtcars) + geom_bar(aes(carb))

library(magrittr)
library(multipanelfigure)

figure1 <- multi_panel_figure(columns = 2, rows = 3, panel_label_type = "upper-roman")

figure1 %<>%
  fill_panel(q1, column = 1, row = 1:3) %<>%
  fill_panel(q2, column = 2, row = 1) %<>%
  fill_panel(q3, column = 2, row = 2) %<>%
  fill_panel(q4, column = 2, row = 3)
#> `geom_smooth()` using method = 'loess' and formula 'y ~ x'
figure1

Created on 2018-07-16 by the reprex package (v0.2.0.9000).