如何将forecast包checkresiduals生成的分面图转为可调用ggplot2对象

要把checkresiduals()生成的残差分面图转化为可后续调用的ggplot2对象，你需要手动提取残差数据，用ggplot2分别绘制三个子图后拼接，具体步骤如下：

1. 准备环境与提取残差数据 #

先确保所需包已安装加载，再从预测对象中提取残差并整理为数据框：

# 如果未安装patchwork包，先执行安装
# install.packages("patchwork")

library(fabletools)
library(forecast)
library(ggplot2)
library(patchwork)

# 生成预测对象
Test_plt <- fdeaths %>% forecast()

# 提取残差并转为带时间索引的数据框
resid_df <- data.frame(
  time = time(residuals(Test_plt)),
  resid = as.numeric(residuals(Test_plt))
)

2. 复刻三个子图 #

checkresiduals()默认生成三个面板，我们逐个用ggplot2实现：

残差时间序列图 #

p1 <- ggplot(resid_df, aes(x = time, y = resid)) +
  geom_line(color = "#0072B2") +
  geom_hline(yintercept = 0, linetype = "dashed", color = "#D55E00") +
  labs(x = "时间", y = "残差", title = "残差时间序列") +
  theme_minimal()

残差直方图+密度曲线 #

p2 <- ggplot(resid_df, aes(x = resid)) +
  geom_histogram(aes(y = after_stat(density)), bins = 15, fill = "#0072B2", alpha = 0.7) +
  geom_density(color = "#D55E00", linewidth = 1) +
  labs(x = "残差", y = "密度", title = "残差分布") +
  theme_minimal()

残差QQ图 #

p3 <- ggplot(resid_df, aes(sample = resid)) +
  stat_qq(color = "#0072B2") +
  stat_qq_line(color = "#D55E00", linewidth = 1) +
  labs(x = "理论分位数", y = "样本分位数", title = "QQ图") +
  theme_minimal()

3. 拼接为分面布局 #

用patchwork复刻原函数的2行1列布局：

# 拼接成最终可调用的ggplot2对象
residual_plot <- (p1 + p2) / p3 +
  plot_layout(heights = c(1, 1))

# 调用图形对象查看
residual_plot

生成的residual_plot是标准ggplot2对象，后续可随时调用、修改主题或添加元素。若需保留原函数的Ljung-Box检验结果，可单独提取：

lb_test <- checkresiduals(Test_plt)
# 查看检验结果
lb_test

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