How to add error bars to barplot in R [closed] - r

Question

Closed. This question needs debugging details. It is not currently accepting answers.
Edit the question to include desired behavior, a specific problem or error, and the shortest code necessary to reproduce the problem. This will help others answer the question.
Closed 5 years ago.
Improve this question
I am new to R and have produced a graph, but I want to add error bars as simply as possible and I do not know how.
ana <- read.table(text="Infiltration Grazing Burn
3301.145496 G S
8165.771889 U S
9937.833576 G L
11576.5892 U L
32739.07643 G N
25923.84328 U N", header=TRUE)
That is my data and below is the code I have used.
barplot(xtabs(ana$Infiltration ~ ana$Grazing + ana$Burn ),beside = TRUE, col = c( "tan4", "darkgreen"), xlab = "Burn Treatment", names = c( "Long Rotation", "Burned 1954", "Short Rotation" ) , ylab = "Mean Infiltration Rate (mm/h) " , legend = c( "Grazed", "Ungrazed"), args.legend = list(title = "Graze Treatment", x = "topright", cex = .7), ylim = c(0, 35000) )
as I am new to R please explain as simply as possible!

Answer 1

This is a basic ggplot2 implementation of what you are after
library(dplyr)
library(ggplot2)
library(magrittr)
## Read in the q data
df <- read.table(text = "Infiltration Grazing Burn
3301.145496 G S
8165.771889 U S
9937.833576 G L
11576.5892 U L
32739.07643 G N
25923.84328 U N",
header = TRUE)
## Add test Lower and upper bounds, trans varnames
df <- df %>%
mutate(ll = Infiltration * 0.9,
hh = Infiltration * 1.1) %>%
mutate(Grazing = Grazing %>%
recode(G = "Grazed", U = "Ungrazed"),
Burn = Burn %>%
recode(S = "Short Rotation", L = "Long Rotation", N = "Burned 194")) %>%
rename(`Graze Treatment` = Grazing)
## Basic boxplot with ci's
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`)) +
geom_bar(stat = "identity", position = "dodge") +
geom_errorbar(aes(ymin = ll, ymax = hh), position = "dodge") +
theme_minimal() +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
It looks like this:
In general boxplots with whiskers are a bit hard to interpret. It might be better to use something like this..
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_point(stat = "identity", position = position_dodge(width = 1), size = 3) +
geom_linerange(aes(ymin = ll, ymax = hh), position = position_dodge(width = 1),
alpha = 0.4, size = 3) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
Note: If you have the raw data from which you generated your confidence intervals you might be better served using a boxplot (with geom_boxplot), a violin plot (with geom_violin) or even a ridge plot (ggridges:geom_density_ridges).
Some possible extensions
If the underlying data is available we can do much better. There are several options, which one you pick comes down to your use case and the size of your data.
First lets generate some sample data.
library(dplyr)
library(ggplot2)
library(tidyr)
library(tibble)
## Read in the q data
df <- read.table(text = "Infiltration Grazing Burn
3301.145496 G S
8165.771889 U S
9937.833576 G L
11576.5892 U L
32739.07643 G N
25923.84328 U N",
header = TRUE)
## Generate and clean some sample data
df <- df %>%
as_tibble %>%
mutate(Infiltration = map(Infiltration, function(x) {
tibble(Infiltration = rnorm(n = 1000,
mean = x,
sd = 0.1 * x),
id = 1:1000)
})) %>%
unnest() %>%
mutate(Grazing = Grazing %>%
recode(G = "Grazed", U = "Ungrazed"),
Burn = Burn %>%
recode(S = "Short Rotation", L = "Long Rotation", N = "Burned 194")) %>%
rename(`Graze Treatment` = Grazing)
Now lets make some plots .
The underlying data with jitter.
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_jitter(position = position_jitterdodge(), alpha = 0.1) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
Boxplots
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_boxplot(alpha = 0.4) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
Violin plots
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_violin(draw_quantiles = c(0.25, 0.5, 0.75), alpha = 0.4) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
Points with mean, 1 and 2 standard deviations
df %>%
group_by(`Graze Treatment`, Burn) %>%
summarise(
mean = mean(Infiltration),
sd = sd(Infiltration),
lll = mean - 2 * sd,
ll = mean - sd,
hh = mean + sd,
hhh = mean + 2*sd) %>%
ggplot(aes(x = Burn, y = mean, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_point(stat = "identity", position = position_dodge(width = 1), size = 3) +
geom_linerange(aes(ymin = lll, ymax = hhh), position = position_dodge(width = 1),
alpha = 0.4, size = 3) +
geom_linerange(aes(ymin = ll, ymax = hh), position = position_dodge(width = 1),
alpha = 0.6, size = 3) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
x = "Infiltration",
caption = "Errorbars represent ....")
Both jittered points and violin plots
df %>%
ggplot(aes(x = Burn, y = Infiltration, fill = `Graze Treatment`, col = `Graze Treatment`)) +
geom_violin(draw_quantiles = c(0.25, 0.5, 0.75),
position = position_dodge(width = 1),
aes(fill = NULL)) +
geom_jitter(position = position_jitterdodge(dodge.width = 1), alpha = 0.01) +
theme_minimal() +
expand_limits(y = 0) +
labs(y = "Mean Infiltration Rate (mm/h)",
caption = "Errorbars represent ....")
and any other summary plot overlayed with the raw data. This falls down when you have lots of data in which case one of the summary plots by itself will be better.