R语言计算quantiles分位差及POE指标可视化代码咨询

首先确保你已将原始数据导入为R数据框，命名为poe_df，列对应为R、Year、Time、POE、VALUE。

方案1：tidyverse生态实现（推荐） #

步骤1：数据预处理 #

# 加载依赖包
library(dplyr)

# 将POE列的百分号格式转为数值
poe_df <- poe_df %>%
  mutate(POE_num = as.numeric(sub("%", "", POE)))

步骤2：分组计算分位数差 #

默认按年份、时间分组计算两种常用分位数差：90%分位范围（95%分位值减10%分位值）、四分位距（75%分位值减25%分位值），可按需调整分组维度和分位数区间。

quantile_diff <- poe_df %>%
  group_by(Year, Time) %>%
  summarise(
    # 90%分位范围
    range_90 = VALUE[POE_num == 95] - VALUE[POE_num == 10],
    # 四分位距IQR
    IQR = VALUE[POE_num == 75] - VALUE[POE_num == 25],
    # 可选保留中位数做参考
    median = VALUE[POE_num == 50],
    .groups = "drop"
  )

# 输出计算结果
print(quantile_diff)

步骤3：分位数差可视化（可选） #

library(ggplot2)

ggplot(quantile_diff, aes(x = Time, group = 1)) +
  geom_ribbon(aes(ymin = median - range_90/2, ymax = median + range_90/2), fill = "lightblue", alpha = 0.3) +
  geom_ribbon(aes(ymin = median - IQR/2, ymax = median + IQR/2), fill = "steelblue", alpha = 0.5) +
  geom_line(aes(y = median), color = "darkblue", linewidth = 1) +
  labs(title = "POE指标分位数差时序表现", y = "POE值", x = "时间") +
  theme_minimal()

方案2：基础R实现（无需额外依赖） #

# 预处理POE列为数值
poe_df$POE_num <- as.numeric(sub("%", "", poe_df$POE))
# 生成唯一分组
groups <- unique(poe_df[,c("Year", "Time")])
# 遍历分组计算分位数差
quantile_diff <- do.call(rbind, lapply(1:nrow(groups), function(i) {
  cur_y <- groups$Year[i]
  cur_t <- groups$Time[i]
  sub_df <- poe_df[poe_df$Year == cur_y & poe_df$Time == cur_t,]
  data.frame(
    Year = cur_y,
    Time = cur_t,
    range_90 = sub_df$VALUE[sub_df$POE_num == 95] - sub_df$VALUE[sub_df$POE_num == 10],
    IQR = sub_df$VALUE[sub_df$POE_num == 75] - sub_df$VALUE[sub_df$POE_num == 25],
    median = sub_df$VALUE[sub_df$POE_num == 50]
  )
}))

注意事项 #

• 若数据包含多个R维度的分组，只需在分组条件中新增R列即可，比如tidyverse方案中改成group_by(R, Year, Time)
• 如需自定义分位数差的区间，修改对应代码块中匹配的POE_num数值即可

内容的提问来源于stack exchange，提问作者Steven