如何判断ggplot2绘制的图表中各数据点所属的特定分区

方法一：空间多边形匹配法（推荐，适配任意形状分区边界） #

无需手动梳理复杂的不等式判断逻辑，通用性极强，后续调整分区边界也不需要重写判断规则。

实现步骤： #

1. 加载依赖包

install.packages("sf")
library(sf)
library(tidyverse)

2. 定义边界函数，采样构造分区多边形

# 定义三条分区边界函数
f1 <- function(x) ((x - 1.05) * 5)^1.0526
f2 <- function(x) ((x - 1.1) * 1.5)^1.0989
f3 <- function(x) x

# 采样x轴坐标点，数值越大分区匹配精度越高
x_seq <- seq(-2, 20, length.out = 1000)

# 生成三条边界的完整坐标
line1 <- tibble(x = x_seq, y = f1(x)) %>% filter(x >= 1.31)
line2 <- tibble(x = x_seq, y = f2(x)) %>% filter(x >= 2.82)
line3 <- tibble(x = x_seq, y = f3(x))

# 构造四个分区的闭合多边形空间对象
zone_sf <- list(
  # CCS区：最下方
  CCS = rbind(
    c(-2, 1),
    line3 %>% as.matrix(),
    c(20, 1),
    c(-2, 1)
  ),
  # SD区：左上方
  SD = rbind(
    c(-2, 1000),
    line3 %>% filter(x <= 2.82) %>% as.matrix(),
    c(-2, 0),
    c(-2, 1000)
  ),
  # CC区：中间区域
  CC = rbind(
    line1 %>% as.matrix(),
    rev(line2 %>% as.matrix()),
    line3 %>% filter(x >= 2.82) %>% arrange(desc(x)) %>% as.matrix(),
    line3 %>% filter(x == 2.82) %>% as.matrix(),
    line1[1,] %>% as.matrix()
  ),
  # TC区：右上方
  TC = rbind(
    c(20, 1000),
    rev(line1 %>% as.matrix()),
    c(20, f1(20)),
    c(20, 1000)
  )
) %>% 
  lapply(st_polygon) %>% 
  st_sfc(crs = NA) %>% 
  st_sf(zone = c("CCS", "SD", "CC", "TC"), geometry = .)

3. 匹配样本点所属分区

# 样本点转为空间点对象
sample_data_sf <- sample_data %>% 
  st_as_sf(coords = c("x", "y"), remove = F)

# 关联得到每个点的分区标签
sample_data_with_zone <- st_join(sample_data_sf, zone_sf)

4. 按分区给点上色绘图

ggplot(
  data = sample_data_with_zone,
  aes(x, y, color = zone)
) +
  geom_point(size = 3) +
  scale_x_continuous(
    expand = c(0, 0),
    minor_breaks = NULL,
    breaks = seq(-2, 20, by = 2)
  ) +
  scale_y_continuous(
    trans = "log10",
    expand = c(0, 0),
    minor = c(seq(1, 10, by = 1), seq(10, 100, by = 10), seq(100, 1000, by = 100))
  ) +
  coord_cartesian(
    xlim = c(-2, 20),
    ylim = c(1, 1000),
    expand = c(0, 0)
  ) +
  geom_function(fun = f1, xlim = c(1.31, 20)) +
  geom_function(fun = f2, xlim = c(2.82, 20)) +
  geom_function(fun = f3, xlim = c(-2, 20)) +
  geom_text(
    data = zoning_data,
    aes(x = x, y = y, label = label),
    inherit.aes = F
  ) +
  theme(aspect.ratio = 1)

方法二：函数判断法（适合边界规则简单的场景） #

如果不想额外安装空间分析包，可以直接根据边界函数写条件判断打标签：

# 定义带定义域校验的边界函数
f1 <- function(x) ifelse(x >= 1.31, ((x - 1.05) * 5)^1.0526, Inf)
f2 <- function(x) ifelse(x >= 2.82, ((x - 1.1) * 1.5)^1.0989, Inf)
f3 <- function(x) x

# 给样本点打分区标签
sample_data_with_zone <- sample_data %>% 
  mutate(
    zone = case_when(
      y < f3(x) ~ "CCS",
      x < 1.31 ~ "SD",
      y < f2(x) ~ "CC",
      y < f1(x) ~ "CC",
      y >= f1(x) ~ "TC"
    )
  )

得到带分区标签的数据后，参考方法一的绘图代码即可实现分色展示。

内容的提问来源于stack exchange，提问作者erdi