问题

我有以下的CSV文件：

group;response
stepOne;107
stepOne;946
stepTwo;184
stepTwo;456
...

我正在将它读入数据框中，计算统计数据，并将该统计数据反映在误差条形图上。以下是代码：

# 加载所需的库
library(ggplot2)
library(dplyr)
library(viridis)

# 读取输入的CSV文件
data <- read.csv("~/Documents/r/input_file.csv", sep=";", header=TRUE)

x_axis_name <- names(data)[2]
y_axis_name <- names(data)[1]
title <- sprintf("置信区间：%s", names(data)[1])

# 为每个组计算均值和置信区间
group_stats <- data %>%
  group_by(group) %>%
  summarise(
    mean = mean(response),
    lower_ci_90 = mean - qt(0.90, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_90 = mean + qt(0.90, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_95 = mean - qt(0.95, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_95 = mean + qt(0.95, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_99 = mean - qt(0.99, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_99 = mean + qt(0.99, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_999 = mean - qt(0.999, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_999 = mean + qt(0.999, df = n() - 1) * (sd(response) / sqrt(n()-1))
  ) %>%
  arrange(desc(group))

# 从group_stats中找到绝对最大值
limit_x_left = 0
limit_x_right <- 1.15*max(unlist(group_stats[, sapply(group_stats, is.numeric)]))

# 生成图形
graph <- ggplot(group_stats, aes(y = group[order(group, decreasing=TRUE)], x = mean, color = group)) +
  geom_errorbarh(aes(xmin = lower_ci_90, xmax = upper_ci_90), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_errorbarh(aes(xmin = lower_ci_95, xmax = upper_ci_95), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_errorbarh(aes(xmin = lower_ci_99, xmax = upper_ci_99), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_point(size = 3) +
  labs(x = x_axis_name, y = y_axis_name, title = title, color = "图例") +
  xlim(limit_x_left, limit_x_right) +
  scale_color_manual(values = group_colors)

# 显示图形
print(graph)

无论我如何初始化group_stats数据框，ggplot都会按字母顺序降序排列条形。试图在ggplot内部重新排序不起作用：

y = group[order(group, decreasing=FALSE)]
或
y = group[order(group, decreasing=TRUE)]
被忽略。

但是，如果我使用
y = group[order(mean, decreasing=TRUE)]
值将按均值完美排序。

所以，有没有办法改变误差条的排序方式？我想按字母升序排列它们。

<details>
<summary>英文:</summary>

I have the following CSV file:
```
group;response
stepOne;107
stepOne;946
stepTwo;184
stepTwo;456
...
```
I am reading it into dataframe, calculate the stats and reflect that stats on graph as errorbarsh. The code is the following

```
# Load required libraries
library(ggplot2)
library(dplyr)
library(viridis)

# Read the input CSV file
data &lt;- read.csv(&quot;~/Documents/r/input_file.csv&quot;, sep=&quot;;&quot;, header=TRUE)

x_axis_name &lt;- names(data)[2]
y_axis_name &lt;- names(data)[1]
title &lt;- sprintf(&quot;Confidence interval for %s&quot;, names(data)[1])

# Calculate mean and confidence intervals for each group
group_stats &lt;- data %&gt;%
  group_by(group) %&gt;%
  summarise(
    mean = mean(response),
    lower_ci_90 = mean - qt(0.90, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_90 = mean + qt(0.90, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_95 = mean - qt(0.95, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_95 = mean + qt(0.95, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_99 = mean - qt(0.99, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_99 = mean + qt(0.99, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    lower_ci_999 = mean - qt(0.999, df = n() - 1) * (sd(response) / sqrt(n()-1)),
    upper_ci_999 = mean + qt(0.999, df = n() - 1) * (sd(response) / sqrt(n()-1))
  ) %&gt;%
  arrange(desc(group))

# Find absolute maximum from group_stats
limit_x_left = 0
limit_x_right &lt;- 1.15*max(unlist(group_stats[, sapply(group_stats, is.numeric)]))

# Generate the graph
graph &lt;- ggplot(group_stats, aes(y = group[order(group, decreasing=TRUE)], x = mean, color = group)) +
  geom_errorbarh(aes(xmin = lower_ci_90, xmax = upper_ci_90), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_errorbarh(aes(xmin = lower_ci_95, xmax = upper_ci_95), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_errorbarh(aes(xmin = lower_ci_99, xmax = upper_ci_99), height = 0.0, linewidth = 2.5, alpha = 0.25) +
  geom_point(size = 3) +
  labs(x = x_axis_name, y = y_axis_name, title = title, color = &quot;Legend&quot;) +
  xlim(limit_x_left, limit_x_right) +
  scale_color_manual(values = group_colors)


# Display the graph
print(graph)
```

No matter how I initially arrange group_stats dataframe, ggplot will put bars ordering them alphabetically DESC. 

Trying to reorder it inside ggplot leads to nothing: 
y = group[order(group, decreasing=FALSE)] 
or
y = group[order(group, decreasing=TRUE)] 
is being ignored.
[![enter image description here][1]][1]

However if I put 
y = group[order(mean, decreasing=TRUE)] the values are being sorted by mean perfectly.
[![enter image description here][2]][2]

So, is there any way to change order of errorbars are being put onto plot? I want to order them alphabetically ASC.


  [1]: https://i.stack.imgur.com/nc69Q.png
  [2]: https://i.stack.imgur.com/0fM9B.png

</details>


# 答案1
**得分**: 1

根据 @jared_mamrot，使用 y = fct_rev(group) 而不是 y = group[order(group, decreasing=TRUE)] 可以解决这个问题。需要安装 forcats 包：

```install.packages("forcats")```

或者安装 tidyverse 包：

```install.packages("tidyverse")```

然后在代码中添加：

```library(forcats)```

感谢 @jared_mamrot。

<details>
<summary>英文:</summary>

Per @jared_mamrot, y = fct_rev(group) instead of y = group[order(group, decreasing=TRUE)] solves the issue. Need to install forcats: 

```install.packages(&quot;forcats&quot;)```
(or ```install.packages(&quot;tidyverse&quot;)```)

and add 

```library(forcats)```

to the code.

Thanks @jared_mamrot

</details>