Align multiple ggplot2 graphs with a common x axis and different y axes, each with different y-axis labels.

plot_aligned_series.R

#' When plotting multiple data series that share a common x axis but different y axes,
#' we can just plot each graph separately. This suffers from the drawback that the shared axis will typically
#' not align across graphs due to different plot margins.
#' One easy solution is to reshape2::melt() the data and use ggplot2's facet_grid() mapping. However, there is
#' no way to label individual y axes.
#' facet_grid() and facet_wrap() were designed to plot small multiples, where both x- and y-axis ranges are
#' shared acros all plots in the facetting. While the facet_ calls allow us to use different scales with
#' the \code{scales = "free"} argument, they should not be used this way.
#' A more robust approach is to the grid package grid.draw(), rbind() and ggplotGrob() to create a grid of 
#' individual plots where the plot axes are properly aligned within the grid.
#' Thanks to https://rpubs.com/MarkusLoew/13295 for the grid.arrange() idea.

library(ggplot2)
library(grid)
library(dplyr)
library(lubridate)

#' Create some data to play with. Two time series with the same timestamp.
df <- data.frame(DateTime = ymd("2010-07-01") + c(0:8760) * hours(2), series1 = rnorm(8761), series2 = rnorm(8761, 100))

#' Create the two plots.
plot1 <- df %>%
  select(DateTime, series1) %>%
  na.omit() %>%
  ggplot() +
  geom_point(aes(x = DateTime, y = series1), size = 0.5, alpha = 0.75) +
  ylab("Red dots / m") +
  theme_minimal() +
  theme(axis.title.x = element_blank())

plot2 <- df %>%
  select(DateTime, series2) %>%
  na.omit() %>%
  ggplot() +
  geom_point(aes(x = DateTime, y = series2), size = 0.5, alpha = 0.75) +
  ylab("Blue drops / L") +
  theme_minimal() +
  theme(axis.title.x = element_blank())

grid.newpage()
grid.draw(rbind(ggplotGrob(plot1), ggplotGrob(plot2), size = "last"))

plot_aligned_series2.R

# To plot two series vertically aligned with only one labelled x-axis,
# we can remove the axes from the top plot and then plot the two graphs
# together using either the egg package or the cowplot package

library(ggplot2) # 3.2.1
library(dplyr)
library(lubridate)
library(cowplot) # 1.0.0
library(egg) # 0.4.5

#' Create some data to play with. Two time series with the same timestamp.
df <- data.frame(DateTime = ymd("2010-07-01") + c(0:8760) * hours(2), 
                 series1 = rnorm(8761), 
                 series2 = rnorm(8761, 100))

#' Create the two plots.
plot1 <- df %>%
  select(DateTime, series1) %>%
  na.omit() %>%
  ggplot() +
  geom_point(aes(x = DateTime, y = series1), size = 0.5, alpha = 0.75) +
  ylab("Red dots / m") +
  theme_minimal() +
  theme(axis.title.x = element_blank(),
        axis.text.x = element_blank())

plot2 <- df %>%
  select(DateTime, series2) %>%
  na.omit() %>%
  ggplot() +
  geom_point(aes(x = DateTime, y = series2), size = 0.5, alpha = 0.75) +
  ylab("Blue drops / L") +
  theme_minimal() +
  theme(axis.title.x = element_blank())

# Draw the two plot aligned vertically, with the top plot 1/3 of the height
# of the bottom plot
cowplot::plot_grid(plot1, plot2, align = "v", ncol = 1, rel_heights = c(0.25, 0.75))
egg::ggarrange(plot1, plot2, heights = c(0.25, 0.75))

@ADam-Z514
It should work the same with grid.arrange. You should have to convert to Grob first and bind them together with rbind.

grid.arrange(rbind(ggplotGrob(graph_weight_ori), ggplotGrob(bpgraph_weight_imp), size = "first"))

This code is great, but is there a way that you can bind both plots together so that x-axis for the upper plot is only separated by a line so as not to show the numbers. Basically, I want two plots directly on top of each other with the same scales for the y-axis but two different variables. Is this possible? I am having trouble finding any code to do this exactly.

for some reason I got this plot to work and now the select command is giving me an error: "Error in select(., d34S, d13C) : unused arguments (d34S, d13C)". The libraries I am running in the background are:

library(tidyverse)
library(ggplot2)
library(broom)
library(sjstats)
library(plotly)
library(fastDummies)
library(MASS)
library(AICcmodavg)
library(car)
library(MASS)
library(grid)
library(dplyr)
library(lubridate)

Can anybody help me?

@bggalvenmon Some package is masking dplyr's select: use dplyr::select

@bggalvenmon Some package is masking dplyr's select: use dplyr::select

I think there is a select function loaded through the MASS package - I recently had a script failing because of this issue.

If you don't want to keep calling the function direction you can add the following to the beginning of your script: select <- dplyr::select

did something like:

p1 <- ggplot(...) + geom_histogram()
p2 <- ggplot(...) + geom_boxplot() + geom_point() + coord_flip()
grid.newpage()
grid.draw(...)

and the axis were close, but not right on.

edit: was able to get them to align by adding:

p1 <- p1 + scale_x_continuous(limits=c(x_min, x_max))
p2 <- p2 + scale_y_continuous(limits=c(x_min, x_max))

thanks!

very nice solution, especially if one wants to compile 2-by-2 histograms from separate data frames.

Thanks!

Great help, thanks very much for posting!!

Many thanks! Solved my issue.

Really nice solution!!

Thank you very much for your great help

If you want to take this:

p1 <- p1 + scale_x_continuous(limits=c(x_min, x_max))
p2 <- p2 + scale_y_continuous(limits=c(x_min, x_max))

a step further, you can do something like:

x_min=min(c(layer_scales(p1)$x$range$range[[1]],layer_scales(p2)$x$range$range[[1]]))
x_max=max(c(layer_scales(p1)$x$range$range[[2]],layer_scales(p2)$x$range$range[[2]]))
p1<-p1+scale_x_continuous(limits=c(x_min, x_max))
p2<-p2+scale_x_continuous(limits=c(x_min, x_max))

If you want to take this:

p1 <- p1 + scale_x_continuous(limits=c(x_min, x_max))
p2 <- p2 + scale_y_continuous(limits=c(x_min, x_max))

a step further, you can do something like:

x_min=min(c(layer_scales(p1)$x$range$range[[1]],layer_scales(plt_mod_cond)$x$range$range[[1]]))
x_max=max(c(layer_scales(plt_mod_all)$x$range$range[[2]],layer_scales(plt_mod_cond)$x$range$range[[2]]))
p1<-p1+scale_x_continuous(limits=c(x_min, x_max))
p2<-p2+scale_x_continuous(limits=c(x_min, x_max))