5

I am trying to generate 10 pairs of plots with a few pairs per page of plots, and am using a for loop to construct the pairs. However, the plots are sent to the device as separate plots instead of pages.

The MWE below has identical constructions for base graphics and ggplot versions, but the base graphics works and ggplot does not. What do I need to do to get the pagination correct in the second version?

library(ggplot2)
attach(mtcars)

# correct configuration
par(mfrow=c(2,2))
for (ii in 1:3){
  vars <- c("wt", "disp", "wt")
  plot(get(vars[ii]), mpg)
  hist(get(vars[ii]))
}

# places each on separate plot
par(mfrow=c(2,2))
for (ii in 1:3){
  vars <- c("wt", "disp", "wt")
  p <- ggplot(mtcars, aes(get(vars[ii]), mpg)) + geom_point(size=4)
  plot(p)
  p <- ggplot(mtcars, aes(get(vars[ii]))) + geom_histogram()
  plot(p)
}

detach(mtcars)
Tung
  • 26,371
  • 7
  • 91
  • 115
JenB
  • 17,620
  • 2
  • 17
  • 45
  • 6
    `par` doesn't operate on `ggplot` graphics, only base R. Do you instead want to facet the plots? – camille Jun 19 '18 at 13:15
  • 1
    An potential alternative to facets would be `gridExtra::grid.arrange`; see e.g. [this post](https://stackoverflow.com/questions/1249548/side-by-side-plots-with-ggplot2). – Maurits Evers Jun 19 '18 at 13:16
  • Good to know that `par` is not available. It's not a `facet` problem - different plots. I will have to see if I can make `gridExtra::grid.arrange` or `cowplot::plot_grid` flexible enough to deal with not naming the individuals plots. – JenB Jun 19 '18 at 13:41
  • You might try storing your plots in a list for work with the packages/functions you mentioned. – aosmith Jun 19 '18 at 14:08

1 Answers1

4

Here is one way to do it with cowplot::plot_grid. The plot_duo function uses tidyeval approach in ggplot2 v3.0.0

# install.packages("ggplot2", dependencies = TRUE)

library(rlang)
library(dplyr)
library(ggplot2)
library(cowplot)

plot_duo <- function(df, plot_var_x, plot_var_y) {

  if (is.character(plot_var_x)) {
    print('character column names supplied, use ensym()')
    plot_var_x <- ensym(plot_var_x)
  } else {
    print('bare column names supplied, use enquo()')
    plot_var_x <- enquo(plot_var_x)
  }

  if (is.character(plot_var_y)) {
    plot_var_y <- ensym(plot_var_y)
  } else {
    plot_var_y <- enquo(plot_var_y)
  }

  pts_plt <- ggplot(df, aes(x = !! plot_var_x, y = !! plot_var_y)) + geom_point(size = 4)
  his_plt <- ggplot(df, aes(x = !! plot_var_x)) + geom_histogram()

  duo_plot <- plot_grid(pts_plt, his_plt, ncol = 2)
}

### use character column names
plot_vars1 <- c("wt", "disp", "wt")
plt1 <- plot_vars1 %>% purrr::map(., ~ plot_duo(mtcars, .x, "mpg"))
#> [1] "character column names supplied, use ensym()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> [1] "character column names supplied, use ensym()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> [1] "character column names supplied, use ensym()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

plot_grid(plotlist = plt1, nrow = 3)


### use bare column names
plot_vars2 <- alist(wt, disp, wt)
plt2 <- plot_vars2 %>% purrr::map(., ~ plot_duo(mtcars, .x, "mpg"))
#> [1] "bare column names supplied, use enquo()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> [1] "bare column names supplied, use enquo()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> [1] "bare column names supplied, use enquo()"
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

plot_grid(plotlist = plt2, nrow = 3)

To separate plots into multiple pages, we can use gridExtra::marrangeGrob

ml1 <- marrangeGrob(plt, nrow = 2, ncol = 1)

# Interactive use
ml1


# Non-interactive use, multipage pdf
ggsave("multipage.pdf", ml1)
Tung
  • 26,371
  • 7
  • 91
  • 115
  • Thank you. This almost works. In my MWE, you can see I am generating 6 plots with 2x2 for the first 'page' and 2x1 on the second. I have adapted your code to my real data, but the `do.call` places all the plots on a single page. Is there a way to force the 2x2 limit so that a separate plot is created for each 4 subplots? – JenB Jun 19 '18 at 15:17