R语言诊断图中平方根标准化残差轴font.lab=2不生效问题

复现代码 #

df <- structure(list(sales1 = c(11.3208, 12.9151, 18.8947, 14.6739, 8.6493, 9.5238, 7.6923, 0.0017, 8.0477, 6.7358, 6.1441, 21.7939, 4.2553, 0.0017, 11.0196, 6.2762, 13.2316, 5.0676, 5.6235, 14.9893 ), store = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 4L, 4L), levels = c("1", "2", "3", "4", "5", "6"), class = "factor"), day = structure(c(1L, 2L, 3L, 4L, 5L, 6L, 1L, 2L, 3L, 4L, 5L, 6L, 1L, 2L, 3L, 4L, 5L, 6L, 1L, 2L), levels = c("1", "2", "3", "4", "5", "6"), class = "factor")), row.names = c(NA, 20L), class = "data.frame")
par(mfrow= c(2,2), mgp=c(2, .7, 0), mai=c(0.7, 0.7, 0.7, 0.1), font.lab = 2)
a1 <- aov(sales1 ~ store + day, df)
plot(a1)

问题现象 #

设置font.lab = 2后，模型诊断图的多数轴标签变为粗体，但左下角子图的y轴标签（平方根标准化残差）未生效。

问题原因 #

这是因为plot.aov()绘制左下角的Scale-Location图时，其y轴标签是通过mtext()手动添加的，没有继承全局par(font.lab)参数。R基础绘图中，部分特殊子图的元素会绕过全局参数，直接使用默认样式渲染。

简便替代方案 #

方案1：修改基础绘图函数的标签逻辑 #

直接重写plot.aov的部分行为，让所有标签都应用粗体设置：

# 重写plot.aov函数，补全Scale-Location图的粗体标签
plot.aov <- function(x, ...) {
  # 调用原生plot.aov绘图
  graphics:::plot.aov(x, ...)
  # 定位到左下角子图区域
  par(fig = c(0, 0.5, 0, 0.5), new = TRUE)
  # 重新绘制粗体y轴标签
  mtext("Square Root of Standardized Residuals", side = 2, line = 2, font = 2, cex = par("cex.lab"))
  # 恢复全局绘图区域设置
  par(fig = c(0,1,0,1), new = FALSE)
}

# 重新运行绘图代码
par(mfrow= c(2,2), mgp=c(2, .7, 0), mai=c(0.7, 0.7, 0.7, 0.1), font.lab = 2)
plot(a1)

方案2：用ggplot2重构诊断图（更灵活可控） #

如果不想修改基础函数，使用ggplot2结合broom包生成诊断图，可完全自定义样式：

library(ggplot2)
library(broom)

# 提取模型诊断数据
diagnostic_data <- augment(a1) %>%
  mutate(
    std_resid = rstandard(a1),
    sqrt_abs_std_resid = sqrt(abs(std_resid)),
    leverage = hatvalues(a1)
  )

# 绘制四个诊断图
ggplot() +
  # 残差-拟合值图
  geom_point(data = diagnostic_data, aes(.fitted, .resid)) +
  geom_hline(yintercept = 0, linetype = 2) +
  labs(x = "Fitted Values", y = "Residuals") +
  # QQ图
  geom_qq(data = diagnostic_data, aes(sample = std_resid)) +
  geom_qq_line(data = diagnostic_data, aes(sample = std_resid)) +
  labs(x = "Theoretical Quantiles", y = "Standardized Residuals") +
  # Scale-Location图
  geom_point(data = diagnostic_data, aes(.fitted, sqrt_abs_std_resid)) +
  geom_smooth(data = diagnostic_data, aes(.fitted, sqrt_abs_std_resid), se = FALSE) +
  labs(x = "Fitted Values", y = "Square Root of Standardized Residuals") +
  # 残差-杠杆图
  geom_point(data = diagnostic_data, aes(leverage, std_resid)) +
  geom_smooth(data = diagnostic_data, aes(leverage, std_resid), se = FALSE) +
  labs(x = "Leverage", y = "Standardized Residuals") +
  # 全局设置：所有轴标签粗体
  theme(axis.title = element_text(face = "bold")) +
  # 分面布局
  facet_wrap(~.panel, ncol = 2, scales = "free")

内容的提问来源于stack exchange，提问作者Ann