Overlay boxplot obtained grouping continuous variable with scatterplot of the original variables

Question

I have a two continuous variables, one of which I have categorized in two groups. I want a scatterplot of the original variables over a boxplot of the categorized variable:

library(ggplot2)
data(cancer, package="survival")
lung$age2 <- lung$age >= 75

## Boxplot
ggplot(lung, aes(x=age2, y=wt.loss)) +
  geom_boxplot()
#> Warning: Removed 14 rows containing non-finite values (stat_boxplot).

## Scatterplot
ggplot(lung, aes(x=age, y=wt.loss)) +
  geom_point()
#> Warning: Removed 14 rows containing missing values (geom_point).

^{Created on 2022-02-11 by the reprex package (v2.0.1)}

I want the two graphs together, with the boxplot drawn at 50 and 70. Is that possible? ggplot2 preferred.

Are you using the `cancer` data or `lung` – akrun Feb 11 '22 at 17:51 — akrun, Feb 11 '22 at 17:51

Ruam Pimentel · Answer 1 · 2022-02-11T19:50:47.517

0

See below the best I could come up with. Which is a bit misleading because some dots from the boxplot get mixed up with the scatterplot. You can change colors. Change the axis, and make it more beautiful for sure.

Package

library(dplyr)

Transform the data

lung <- lung %>% 
  mutate(age2 = case_when(age >= 75 ~ 70,
                          age < 75 ~ 50))

Plot

ggplot(lung) +
  geom_boxplot(aes(x=factor(age2), y=wt.loss)) +
  geom_point(aes(x=factor(age), y=wt.loss))

edited Feb 11 '22 at 19:50

answered Feb 11 '22 at 18:01

Ruam Pimentel

1,288
4
16

I'd really need the x-axis to be numeric so that I can also add a smooth line etc. I found an acceptable approximation using tips from this post: https://stackoverflow.com/questions/27050802/how-to-create-geom-boxplot-with-large-amount-of-continuous-x-variables – Claudio Feb 12 '22 at 09:52

score 0 · Answer 2 · answered Feb 12 '22 at 10:06

I found this post that shows how to create a boxplot with a numeric x-axis. That is an acceptable solution to my problem, although not perfect:

library(ggplot2)
data(cancer, package="survival")
lung$age2 <- ifelse(lung$age < 75, 50, 70)
ggplot(lung) +
  geom_boxplot(aes(x=age, y=wt.loss, group=age2)) +
  geom_point(aes(x=age, y=wt.loss))
#> Warning: Removed 14 rows containing non-finite values (stat_boxplot).
#> Warning: Removed 14 rows containing missing values (geom_point).

^{Created on 2022-02-12 by the reprex package (v2.0.1)}

Compared to the optimal solution, I couldn't get the boxes at fixed positions and of the same width (it is proportional to the number of observation). The position of the boxes make sense, however, because it is sensible to put them around the middle value of the corresponding group. I'd like to reduce the box width, though, but the width option seems to have no effect.

Overlay boxplot obtained grouping continuous variable with scatterplot of the original variables

2 Answers2