【发布时间】：2019-12-04 17:19:42
【问题描述】：
我正在为一些李克特数据（沿these lines）制作发散条形图。客户已请求每个组的“平均响应”，将李克特响应视为连续整数（“非常不同意”= 1，“不同意”= 2 等）；这些均值显示在条形顶部的“中性”区域中。
为了透明度，我想将每个李克特响应的数值添加到图例中。我可以将数字添加到标签中（例如，“非常同意（5）”），但我更愿意将它放在颜色框的顶部（例如，在蓝色方块的顶部，表示“非常同意”）。
下面是产生发散条形图的代码：
library(dplyr)
library(ggplot2)
library(RColorBrewer)
# The data.
df = structure(list(group = structure(c(1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L),
.Label = c("Group A", "Group B", "Group C"),
class = "factor"),
response = c(1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 4L, 4L, 4L, 5L, 5L, 5L),
n = c(4, 8, 25, 8, 25, 29, 29, 35, 28, 25, 22, 12, 34, 10, 6),
mean.response = c(3.8, 3, 2.5, 3.8, 3, 2.5, 3.8, 3, 2.5, 3.8, 3, 2.5, 3.8, 3, 2.5),
response.fill = c("#CA0020", "#CA0020", "#CA0020",
"#F4A582", "#F4A582", "#F4A582",
"#F7F7F7", "#F7F7F7", "#F7F7F7",
"#92C5DE", "#92C5DE", "#92C5DE",
"#0571B0", "#0571B0", "#0571B0"),
n.to.plot = c(4, 8, 25, 8, 25, 29, 14.5, 17.5, 14, 25, 22, 12, 34, 10, 6)),
class = c("grouped_df", "tbl_df", "tbl", "data.frame"),
row.names = c(NA, -15L),
groups = structure(list(group = structure(1:3, .Label = c("Group A", "Group B", "Group C"),
class = "factor"),
.rows = list(c(1L, 4L, 7L, 10L, 13L),
c(2L, 5L, 8L, 11L, 14L),
c(3L, 6L, 9L, 12L, 15L))),
row.names = c(NA, -3L),
class = c("tbl_df", "tbl", "data.frame"),
.drop = TRUE))
# Groups, responses, and colors.
n.groups = 3
groups = paste("Group", LETTERS[1:n.groups])
likert.responses = c("Strongly disagree", "Disagree", "Neutral", "Agree", "Strongly agree")
pal = brewer.pal(length(likert.responses), "RdBu")
# Make the plot.
ggplot(data = df, aes(x = group, y = n.to.plot, fill = response.fill)) +
# Start with the "agree" responses.
geom_bar(data = df %>% filter(response >= 3),
stat = "identity") +
# Add the "disagree" responses going the opposite way.
geom_bar(data = df %>%
filter(response <= 3) %>%
mutate(n.to.plot = n.to.plot * -1),
stat = "identity") +
# Add text labels with the mean response for each group.
geom_text(data = df %>%
dplyr::select(group, mean.response) %>%
distinct(),
aes(x = group, y = 0,
label = format(mean.response, nsmall = 1),
fill = NA)) +
# Specify fill colors.
scale_fill_identity("Response", breaks = pal, labels = likert.responses,
guide = "legend") +
# Adjust axis labels.
scale_x_discrete("") +
scale_y_continuous("Number of responses") +
# Swap x and y axes.
coord_flip() +
# Add the prompt text as the title.
ggtitle("I like program XYZ.")
这是我想要的输出：
从this answer 中汲取灵感，我尝试在填充图例中添加label 美学，但没有任何效果：
+ guides(fill = guide_legend(override.aes = list(label = "foo")))
我知道我可以customize the shape of the legend symbols，但问题是我想要两件事：一个有颜色的正方形，以及一个叠加在正方形上的黑色数字。
更新：自定义注释
@M-- 建议使用annotation_custom，如here 所述。为此，我需要弄清楚图例中的颜色框在哪里。这就是我卡住的地方；我可以找到这些盒子的 grobs，但我不知道如何在它们上面放置文本。
寻找其中一个颜色框（将上面的图保存为g；在this answer 的指导下）：
gt = ggplot_gtable(ggplot_build(g))
gb = which(gt$layout$name == "guide-box")
box.grob = gt$grobs[[gb]]$grobs[[1]]$grobs[[3]]
box.grob$x和box.grob$y都是0.5npc；我尝试使用geom_text_npc 添加标签，但标签就在图的中间。显然，我没有正确识别颜色框的位置（或者我没有将其转换为正确绘制坐标）。
library(ggpmisc)
g + geom_text_npc(aes(npcx = 0.5, npcy = 0.5, label = "foo"))
【问题讨论】：
-
我意识到我们没有
response.fill和pal对于breaks来说是不够的！这一切都归结为提供一个可重现的例子。 -
response.fill是df的一部分，pal直接在df下定义。我现在看到我忽略了包含所需的库（RColorBrewer对pal至关重要）；我已经编辑了帖子以添加这些内容。