GGplot Color Outline - r

Question

data(mtcars)
library(ggplot2)
ggplot(mtcars, aes(x = reorder(row.names(mtcars), mpg), y = mpg, fill = factor(cyl))) +
geom_bar(stat = "identity")
This will ggplot the bars with solid fills but what if I wish to use the same fill colors as outlines for some measures but solid fills for others. For example if 'am' equals to 1 it is solid fill but if 'am' equals to 0 than it is just an outline fill like this sample:

Answer 1

One option to remove the fill based on a logical condition is to change those values to NA.
library(tidyverse)
d <- head(mtcars) %>%
rownames_to_column() %>%
# make a new variable for fill
# note: don't use ifelse on a factor!
mutate(cyl_fill = ifelse(am == 0, NA, cyl),
# now make them factors
# (you can do this inside ggplot, but that is messy)
cyl = factor(cyl),
cyl_fill = factor(cyl_fill, levels = levels(cyl)))
# plot
p <- ggplot(d) +
aes(x = rowname,
y = mpg,
color = cyl,
fill = cyl_fill
) +
geom_bar(stat = "identity") +
theme(axis.text.x = element_text(angle = 90))
# change the fill color of NA values
p + scale_fill_discrete(drop=FALSE, na.value="white")
If you want NA fill values to be empty and omitted from the legend:
# omit the fill color of NA values
# note: drop=FALSE is still needed to keep the fill and (outline) color values the same
p + scale_fill_discrete(drop=FALSE, na.translate = F)
You can change the color of the outline in the same way (e.g. cyl_color = ifelse(am != 0, NA, Cyl)), but if you want to specify a color like white or black, it will (should) appear in the legend. You can try to hack your way around these wise defaults by plotting non-aesthetic layers behind your main layers, but it usually gets ugly:
head(mtcars) %>%
rownames_to_column() %>%
mutate(cyl_fill = ifelse(am == 0, NA, cyl),
cyl_color = ifelse(am != 0, NA, cyl),
cyl = factor(cyl),
cyl_fill = factor(cyl_fill, levels = levels(cyl)),
cyl_color = factor(cyl_color, levels = levels(cyl))) %>%
ggplot() +
aes(x = rowname,
y = mpg,
color = cyl_color,
fill = cyl_fill
) +
geom_bar(stat = "identity", color = "black") + # NON-AES LAYER FIRST
geom_bar(stat = "identity") + # Covers up the black except where omitted
theme(axis.text.x = element_text(angle = 90))+
scale_fill_discrete(drop=FALSE, na.translate = F) +
scale_color_discrete(drop=FALSE, na.translate = F)

Answer 2

You could assign the desired colors to each level of the fill and color variables. For example:
library(tidyverse)
mtcars %>%
rownames_to_column() %>%
arrange(mpg) %>%
mutate(rowname=factor(rowname, levels=rowname)) %>%
ggplot(aes(x = rowname, y = mpg, fill = factor(am), colour=factor(cyl))) +
geom_col(size=1) +
scale_fill_manual(values=c("0"="white", "1"="red")) +
scale_color_manual(values=c("4"="blue", "6"="orange", "8"="white")) +
theme_classic() +
theme(axis.text.x=element_text(angle=-90, vjust=0.5, hjust=0))

Answer 3

May be, we can do
library(dplyr)
library(ggplot2)
mtcars %>%
mutate(new = case_when(am == 1 ~ factor(cyl)),
new1 = case_when(am !=1 ~ factor(cyl))) %>%
ggplot(aes(x = reorder(row.names(mtcars), mpg), y = mpg,
fill = new, color = new1)) +
geom_bar(stat = 'identity') +
scale_fill_discrete(na.value= NA) + # similar to Devin Judge-Lord post
theme_classic() +
theme(axis.text.x=element_text(angle=-90, vjust=0.5, hjust=0))