plot one set of columns against another set using ggplot2

Question

Can ggplot2 be used to create a plot matrix of one set of columns against another set?

For example, with the dataframe below plot all columns beginning with 'x' against all columns beginning with 'y', to produce a grid of plots.

require("tidyverse")

df <- tibble(
  x1 = sample(10),
  x2 = sample(10),
  x3 = sample(10),
  y1 = sample(10),
  y2 = sample(10)
)

And what if, unlike in the example above, the columns are not named in a regular pattern - is there a way that arbitrary sets of columns can be chosen?

Thanks in advance

Look at the expand.grid function. It should create all 2-way combinations of an arbitrary set of 2 list names. (It's not a part of tidyverse.) — IRTFM
– IRTFM, Commented Jan 12, 2018 at 4:34

Axeman · Accepted Answer · 2018-01-12 10:35:11Z

You can reshape with tidyr::gather then facet:

df_long <- df %>% 
  gather(x_axis, x, contains("x")) %>% 
  gather(y_axis, y, contains("y"))

# A tibble: 60 x 4
   x_axis     x y_axis     y
    <chr> <int>  <chr> <int>
 1     x1    10     y1     6
 2     x1     6     y1    10
 3     x1     5     y1     3
 4     x1     7     y1     8
 5     x1     8     y1     2
 6     x1     1     y1     1
 7     x1     3     y1     5
 8     x1     9     y1     9
 9     x1     4     y1     7
10     x1     2     y1     4
# ... with 50 more rows

Instead of contains you can use any other tidyverse selection function, or just supply the raw column names.

Then plot:

ggplot(df_long, aes(x, y)) + 
    geom_point() + 
    facet_grid(y_axis ~ x_axis, switch = "both") +
    labs(x = NULL, y = NULL) +
    theme(strip.placement = "outside", strip.background = element_blank())

If you need free scales, you can wrap instead:

ggplot(df_long, aes(x, y)) + 
    geom_point() + 
    facet_wrap(~ interaction(y_axis, x_axis), scales = "free")

Claus Wilke · Accepted Answer · 2018-01-12 04:56:11Z

The simplest approach is probably to just loop over all possible combinations, make the respective plot, and then combine all into a grid.

require("tidyverse")

df <- tibble(
  x1 = sample(10),
  x2 = sample(10),
  x3 = sample(10),
  y1 = sample(10),
  y2 = sample(10)
)


group1 <- c("x1", "x2", "x3") # set of variables along x axis
group2 <- c("y1", "y2") # set of variables along y axis

plotlist <- list()
for (x in group1) {
  for (y in group2) {
    p <- ggplot(df, aes_string(x, y)) + geom_point() + ggtitle(paste0(y, " versus ", x))
    plotlist <- append(plotlist, list(p))
  }
}

cowplot::plot_grid(plotlist = plotlist)

The last step here uses the cowplot package, which I wrote. Alternatively, you could use ggarrange from the egg package to place the plots into a grid.

pyg · Accepted Answer · 2018-01-17 11:57:02Z

1

For the sake of completeness, here's a solution that uses ggduo from the GGally package (a function I just became aware of)

require(GGally)    
df %>% ggduo(columnsX = 1:3, columnsY = 4:5)

answered Jan 17, 2018 at 11:57

pyg

84210 silver badges20 bronze badges

Collectives™ on Stack Overflow

plot one set of columns against another set using ggplot2

3 Answers 3

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related