R 数据可视化 —— ggplot 线图

前言

ggplot2 包含很多绘制线条的函数：大致可分为如下几类：

• 连接线：折线(geom_line)、路径线(geom_path)、阶梯线(geom_step)

• 参考线：水平线(geom_hline)、竖直线(geom_vline)、斜线(geom_abline)

• 线段和曲线: geom_segment、geom_spoke、geom_curve

• 函数曲线: geom_function、stat_function

示例

1. 连接线

主要有三种连接线：

1. geom_path：按照它们在数据中出现的顺序连接起来
2. geom_line：按 x 轴上变量的顺序连接起来
3. geom_step：创建一个阶梯图，突出显示数据的变化

常用参数：

• linetype：线条类型
• size：线条大小
• lineend：线端点样式：round, butt, square
• linejoin：线连接点样式：round, mitre, bevel
• arrow：使用 grid::arrow() 函数设置箭头样式

绘制一条简单的时间序列折线

ggplot(economics, aes(date, unemploy)) + geom_line()

绘制多条折线

economics_long %>% 
  subset(variable %in% c("uempmed", "unemploy")) %>%
  ggplot(aes(date, value01, colour = variable)) +
  geom_line()

翻转线条

ggplot(economics, aes(unemploy, date)) + geom_line(orientation = "y")

如果我们更加关注 y 值的变化情况，可以使用 geom_step 绘制阶梯图

recent <- economics[economics$date > as.Date("2013-01-01"), ]

p1 <- ggplot(recent, aes(date, unemploy)) + geom_line()

p2 <- ggplot(recent, aes(date, unemploy)) + geom_step()

plot_grid(p1, p2)

而 geom_path 可以让你探索两个变量是如何随着时间的推移而发生变化的

例如，失业率和个人储蓄率随时间的关系

esamp <- sample_n(economics, 10)
  
m <- ggplot(esamp, aes(unemploy/pop, psavert))

p1 <- m + geom_path()

p2 <- m + geom_path(aes(colour = as.numeric(date)))

plot_grid(p1, p2)

设置箭头

c <- ggplot(economics, aes(x = date, y = pop))
c1 <- c + geom_line(arrow = arrow())

c2 <- c + geom_line(
  arrow = arrow(angle = 15, ends = "both", type = "closed")
)
plot_grid(c1, c2)

更改连接线及端点样式

base <- tibble(x = 1:3, y = c(4, 1, 9)) %>%
  ggplot(aes(x, y))

b1 <- base + geom_path(size = 8)

b2 <- base + geom_path(size = 8, lineend = "round")

b3 <- base + geom_path(size = 8, lineend = "round", colour = "red")

b4 <- base + geom_path(size = 8, linejoin = "mitre", lineend = "butt")

plot_grid(b1, b2, b3, b4)

当线条的中间有 NA 值时，则会有一个断点

df <- data.frame(x = 1:5, y = c(1, 2, NA, 4, 5))
ggplot(df, aes(x, y)) + geom_point() + geom_line()

设置线条类型

economics_long %>% 
  subset(variable %in% c("uempmed", "unemploy")) %>%
  ggplot(aes(date, value01, colour = variable)) +
  geom_line(aes(linetype = factor(variable))) +
  scale_linetype_manual("variable", values = c(5, 3))

注意：无法同时设置渐变色与线条类型，下面的代码将会报错

economics_long %>% 
  subset(variable %in% c("uempmed", "unemploy")) %>%
  ggplot(aes(date, value01, group = variable)) +
  geom_line(aes(colour = value01), linetype = 2)

2. 参考线

为图形添加参考线对图形的注释非常有用，主要有水平、竖直和对角线三种参考线，对应于三个函数：

• geom_hline: yintercept（y 轴截距）
• geom_vline: xintercept（x 轴截距）
• geom_abline: slope（斜率） 和 intercept（截距）

p <- ggplot(mtcars, aes(wt, mpg)) + geom_point()

# 使用固定值
p1 <- p + geom_vline(xintercept = 5)
# 使用向量
p2 <- p + geom_vline(xintercept = 1:5)
# 水平线
p3 <- p + geom_hline(yintercept = 20)
# 斜线
p4 <- p + geom_abline(intercept = 31, slope = -5)

plot_grid(p1, p2, p3, p4)

计算拟合曲线的截距和斜率，然后绘制直线

> coef(lm(mpg ~ wt, data = mtcars))
(Intercept)          wt 
  37.285126   -5.344472 
> p + geom_abline(intercept = 37, slope = -5)

更简单的方式是使用 geom_smooth 绘制拟合直线

p + geom_smooth(method = "lm", se = FALSE)

在绘制分面图形的时候，可以为不同的分面绘制不同的直线

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point() +
  facet_wrap(~ cyl)

mean_wt <- data.frame(cyl = c(4, 6, 8), wt = c(2.28, 3.11, 4.00))
p + geom_hline(aes(yintercept = wt), mean_wt)

也可以添加其他属性

ggplot(mtcars, aes(mpg, wt, colour = wt)) +
  geom_point() +
  geom_hline(aes(yintercept = wt, colour = wt), mean_wt) +
  facet_wrap(~ cyl)

3. 线段和曲线

geom_segment 用于绘制两个点之间的直线，geom_curve 用于绘制两点的曲线。

两个点通过四个参数 (x, y) 和 (xend, yend) 指定坐标。

例如，在散点图中标注两点之间的连接线

b <- ggplot(mtcars, aes(wt, mpg)) +
  geom_point()

df <- data.frame(x1 = 2.320, x2 = 3.520, y1 = 22.8, y2 = 15.5)
b +
  geom_curve(aes(x = x1, y = y1, xend = x2, yend = y2, colour = "curve"), data = df) +
  geom_segment(aes(x = x1, y = y1, xend = x2, yend = y2, colour = "segment"), data = df)

设置不同的曲率

b1 <- b + geom_curve(aes(x = x1, y = y1, xend = x2, yend = y2), data = df, curvature = -0.2)

b2 <- b + geom_curve(aes(x = x1, y = y1, xend = x2, yend = y2), data = df, curvature = 0.9)

plot_grid(b1, b2)

添加箭头

b + geom_curve(
  aes(x = x1, y = y1, xend = x2, yend = y2),
  data = df,
  arrow = arrow(length = unit(0.05, "npc"))
)

使用 geom_segment 通过设置线段大小来绘制直方图

counts <- as.data.frame(table(x = rpois(100,5)))
counts$x <- as.numeric(as.character(counts$x))

ggplot(counts, aes(x, Freq)) +
  geom_segment(aes(xend = x, yend = 0), size = 10, lineend = "butt")

而 geom_spoke 是由坐标点 (x, y) 以及角度 (angle) 和 半径 (radius) 指定的线段

df <- expand.grid(x = 1:10, y=1:10)
df$angle <- runif(100, 0, 2*pi)
df$speed <- runif(100, 0, sqrt(0.1 * df$x))

ggplot(df, aes(x, y)) +
  geom_point() +
  geom_spoke(aes(angle = angle), radius = 0.5)

看起来像是散落的大头针一样

设置可变的半径

ggplot(df, aes(x, y)) +
  geom_point() +
  geom_spoke(aes(angle = angle, radius = speed))

4. 函数曲线

使用 geom_function 或 stat_function 可以绘制指定函数的曲线，例如

set.seed(2021)
ggplot(data.frame(x = rnorm(100)), aes(x)) +
  geom_density() +
  geom_function(fun = dnorm, colour = "red")

绘制了函数在数据范围内的曲线

也可以只指定范围，来绘制无数据的函数曲线

base <- ggplot() + xlim(-5, 5)
base + geom_function(fun = dnorm)

设置函数的参数值

base + geom_function(fun = dnorm, args = list(mean = 2, sd = .5))

其底层原理是在一些离散点上执行函数，然后用线将各函数值连接起来

b1 <- base + stat_function(fun = dnorm, geom = "point")

b2 <- base + stat_function(fun = dnorm, geom = "point", n = 20)

plot_grid(b1, b2)

下面两行代码效果是一样的

b1 <- base + geom_function(fun = dnorm, n = 20)

b2 <- base + stat_function(fun = dnorm, geom = "line", n = 20)

plot_grid(b1, b2)

自定义函数

# 一张图绘制不同的函数
p1 <- base +
  geom_function(aes(colour = "normal"), fun = dnorm) +
  geom_function(aes(colour = "t, df = 1"), fun = dt, args = list(df = 1))
# 使用匿名函数
p2 <- base + geom_function(fun = function(x) 0.5*exp(-abs(x)))
# 同上
p3 <- base + geom_function(fun = ~ 0.5*exp(-abs(.x)))
# 使用自定义函数，效果同上
f <- function(x) 0.5*exp(-abs(x))
p4 <- base + geom_function(fun = f)

plot_grid(p1, p2, p3, p4)

样式图

1. 路线图

sample_n(mtcars, 10) %>%
  ggplot(aes(mpg, disp)) +
  geom_point(colour = "#69b3a2", na.rm = TRUE) +
  geom_segment(aes(xend = c(tail(mpg, n=-1), NA),
                   yend = c(tail(disp, n=-1), NA)),
               arrow = arrow(length=unit(0.3,"cm")),
               colour = "#69b3a2") +
  geom_text(aes(label = disp), hjust = 1.2) +
  theme_bw()

2. 坡度图

library(ggrepel)

mpg %>% 
  group_by(year, manufacturer) %>%
  summarise(value = sum(displ)) %>%
  pivot_wider(names_from = year, values_from = value) %>%
  mutate(class = if_else((`1999` - `2008`) > 0, "#8dd3c7", "#bebada")) %>%
  ggplot() +
  geom_segment(aes(x = 1, xend = 2, y = `1999`, yend = `2008`, colour = class),
               size = .75, show.legend = FALSE) +
  geom_vline(xintercept = 1, linetype = "solid", size = 1, colour = "#ff7f00") +
  geom_vline(xintercept = 2, linetype = "solid", size = 1, colour = "#1f78b4") +
  geom_point(aes(x = 1, y = `1999`), size = 3, shape = 21, fill = "green") +
  geom_point(aes(x = 2, y = `2008`), size = 3, shape = 21, fill = "red") +
  scale_colour_manual(labels = c("Up", "Down"), values = c("#8dd3c7", "#bebada")) +
  xlim(.5, 2.5) +
  
  geom_text_repel(aes(x = 1, y = `1999`, label = `1999`), 
                  hjust = "left", size = 3.5) +
  geom_text_repel(aes(x = 2, y = `2008`, label = `2008`), 
                  hjust = "right", size = 3.5) +
  geom_text(aes(y = 1.03*max(max(`1999`), max(`2008`))), label = "1999", x = 1,
            size = 5, hjust = 1.2) +
  geom_text(aes(y = 1.03*max(max(`1999`), max(`2008`))), label = "2008", x = 2,
            size = 5, hjust = -.2) +
  theme_void()

在这个例子中，由于点有重叠的现象，导致标签也会重叠在一起。

所以我们使用了 ggplot2 的扩展包 ggrepel 的 geom_text_repel 来绘制不重叠标签。