使用R的forecast包为分解时序图添加文本及重命名面板的问题

1. 仅在趋势面板添加竖线旁文本 #

forecast::autoplot(decomposed_ts)生成的分面图，分面变量是Series，取值包括"data"、"trend"、"seasonal"、"random"。要只在趋势面板添加文本，可通过构造标注数据集，配合geom_text指定仅匹配trend面板：

library(forecast)    
library(ggplot2)

data("AirPassengers")
decomposed_ts <- decompose(AirPassengers)

# 构造标注数据，指定分面为trend，设置文本位置和内容
annotation_df <- data.frame(
  Series = "trend",  # 匹配趋势面板的分面变量值
  x = 1954.5,        # 竖线旁的x坐标，略偏移避免重叠
  y = 300,           # 文本的y坐标，根据趋势线高度调整
  label = "关键年份1954"
)

forecast::autoplot(decomposed_ts) + 
  geom_vline(xintercept = 1954, colour = "red") +
  geom_text(data = annotation_df, aes(x = x, y = y, label = label), 
            colour = "red", inherit.aes = FALSE)

注：inherit.aes = FALSE用于避免继承原绘图的美学映射，确保仅使用标注数据框中的坐标参数。

2. 修改面板名称为首字母大写 #

方法一：修改绘图数据的因子水平 #

先将绘图对象保存为变量，再修改其数据中Series列的因子标签：

p <- forecast::autoplot(decomposed_ts) + 
  geom_vline(xintercept = 1954, colour = "red")

# 修改因子水平标签
levels(p$data$Series) <- c("Data", "Trend", "Seasonal", "Random")

print(p)

方法二：使用facet_wrap的labeller参数 #

直接在绘图时覆盖分面标签：

forecast::autoplot(decomposed_ts) + 
  geom_vline(xintercept = 1954, colour = "red") +
  facet_wrap(~Series, scales = "free_y", 
             labeller = as_labeller(c(data = "Data", 
                                      trend = "Trend", 
                                      seasonal = "Seasonal", 
                                      random = "Random")))

两种方法都能将data面板名改为Data，同时可统一调整其他面板的首字母大写格式。

内容的提问来源于stack exchange，提问作者juliefoxsteven