multi-panel scatterplot in Cars package

Question

I'm trying to make a 2 panel figure in RStudio. Simple enough usually:

par(mfrow=c(1,2)) #1*2 plotting window

However, when I make the plot using the scatterplot() function from the car package it seems to override this division of the plotting window.

Reproducible example: The format I'm looking for can be produced with:

par(mfrow=c(1,2))
plot(iris$Sepal.Length,iris$Sepal.Width)
plot(iris$Petal.Width,iris$Sepal.Width)

But I'd like to use scatterplot() for a few reasons. But when I try the same formatting trick again, it doesn't work. Try this block:

library(car)
par(mfrow=c(1,2))
scatterplot(Sepal.Length~Sepal.Width|Species,data=iris,grid="FALSE", boxplots="", reg.line="FALSE",pch=c(0,1,2))
scatterplot(Petal.Width~Sepal.Width|Species,data=iris,grid="FALSE", boxplots="", reg.line="FALSE",pch=c(0,1,2))

Is anyone aware of a workaround for this?

I also tried an alternative structuring with:

m<-rbind(c(1,2))
layout(m)

No luck.

UPDATE This question is largely a duplicate of this one, except I'm looking for a workaround for it rather than the responses left on that page which basically just acknowledge that scatterplot() does in fact override par(mfrow)

Answer 1

You can output side-by-side figures in an rmarkdown document using the chunk argument fig.show='hold'. Another option is to use ggplot2 to create the plot, instead of the cars scatterplot function. I show both approaches below.

Side-by-Side cars::scatterplot plots in a rmarkdown document

Below is an example with PDF output. out.width='3in' sets the actual size of each plot in the output document, regardless of fig.height and fig.width. But you can still adjust fig.height and fig.width to adjust the aspect ratio and the text size relative to the plot area.

---
title: "Untitled"
author: "eipi10"
date: "November 23, 2016"
output: pdf_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r fig.show='hold', out.width='3in', fig.height=5, fig.width=5}
library(car)

scatterplot(Sepal.Length~Sepal.Width|Species,data=iris,grid="FALSE",
            boxplots="", reg.line="FALSE", pch=c(0,1,2))
scatterplot(Petal.Width~Sepal.Width|Species,data=iris,grid="FALSE", 
            boxplots="", reg.line="FALSE", pch=c(0,1,2))
```

Side-by-Side scatterplots using ggplot2

If you're willing to go with ggplot2 then you can get the two-plot layout relatively easily:

library(ggplot2)
library(gridExtra)
theme_set(theme_bw())

grid.arrange(
  ggplot(iris, aes(Sepal.Width, Sepal.Length, colour=Species)) +
    geom_point() +
    geom_smooth(alpha=0.2) +
    theme(legend.position="top"),
  ggplot(iris, aes(Sepal.Width, Petal.Width, colour=Species)) +
    geom_point() +
    geom_smooth(alpha=0.2) +
    theme(legend.position="top"),
  ncol=2)

Customize ggplot2 plots to look more like cars::scatterplot output

You can customize the code above in other ways. If you don't want the confidence bands, add se=FALSE to geom_smooth. If you want different shapes for each species, add aes(shape=Species) to geom_point. If you want the specific shapes used in base graphics, add + scale_shape_manual(values=0:2), etc. You can also get a single legend with a little extra work.

In the code below, I've added these and other customizations to reproduce something closer to your original base graphics plot.

# Components we'll reuse for both plots
my_theme = list(geom_point(aes(shape=Species)),
                geom_smooth(se=FALSE, show.legend=FALSE, lwd=0.8),
                scale_shape_manual(values=0:2),
                scale_colour_manual(values=c("black", "red","green")),
                theme_bw(),
                theme(panel.grid.major=element_blank(),
                      panel.grid.minor=element_blank(),
                      legend.position="top"))

p1 = ggplot(iris, aes(Sepal.Width, Sepal.Length, colour=Species)) +
  my_theme +
  labs(x="Sepal Width", y="Sepal Length") +
  scale_y_continuous(limits=c(3,8)) +
  scale_x_continuous(limits=c(1,5)) 

p2 = ggplot(iris, aes(Sepal.Width, Petal.Width, colour=Species)) +
  my_theme +
  labs(x="Sepal Width", y="Petal Width")

# Function to extract legend
# https://github.com/hadley/ggplot2/wiki/Share-a-legend-between-two-ggplot2-graphs
g_legend<-function(a.gplot){
  tmp <- ggplot_gtable(ggplot_build(a.gplot))
  leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
  legend <- tmp$grobs[[leg]]
  return(legend)}

leg = g_legend(p1)

grid.arrange(leg,
  arrangeGrob(grobs=lapply(list(p1,p2), function(p) p + guides(colour=FALSE, shape=FALSE)), ncol=2), 
  ncol=1, heights=c(1,10))