使用facet_grid2自由刻度时如何保持箱线图图案一致性？

问题 #

我有一个包含两个实验区域不同处理组质量的数据集：其中一个区域有2种处理，另一个区域有4种处理，且两个区域的数值差异较大。我希望通过facet_grid2()（来自ggh4x包）绘制图形，实现两个需求：

• 离散x轴采用自由刻度，使两个分面中的箱线图尺寸一致，分面宽度随处理数量变化；
• 连续y轴采用自由刻度，适配区域间的数值差异。

目前已通过ggh4x实现上述刻度设置，但发现设置x轴自由刻度后，较窄分面中箱线图的条纹间距变小——即使手动指定了图案参数，需要保持箱线图的图案（条纹间距等）在两个分面中一致。

当前代码：

# generate data
treat1 <- c(rep("bare",10), rep("veg",10), rep("bare",10), rep("veg",10), rep("bare",10), rep("veg",10))
treat2 <- c(rep("wet",10), rep("wet",10), rep("wet",10), rep("wet",10), rep("dry",10), rep("dry",10))
area <- c(rep("down",10), rep("down",10), rep("up",10), rep("up",10), rep("up",10), rep("up",10))
weight <- c(runif(10, 20, 50), runif(10, 1, 15), runif(10, 1, 10), runif(10, 1, 5), runif(10, 1, 10), runif(10, 1, 5))

dat_exp <- tibble("treat1" = treat1, "treat2" = treat2, "area" = area, "mass" = weight) %>%
  mutate(treat2_area = paste(treat2, area, sep = "."),
         treat1_treat2 = paste(treat1, treat2, sep = "."))


# plot it

ggplot(data = dat_exp, aes(x= treat1_treat2, y = mass, fill = treat2_area, pattern = treat1)) +
  theme_bw() +
  geom_boxplot_pattern(position = position_dodge(preserve = "single"), color = "black", pattern_colour = "white", pattern_fill = "white", pattern_angle = 45, pattern_density = 0.3, pattern_spacing = 0.025, pattern_key_scale_factor = 0.6) +
  scale_pattern_manual(values = c(bare = "stripe", veg = "none")) +
  ggh4x::facet_grid2(. ~ area, space = "free_x", scales = "free", independent = "y", switch = "both")

问题现象：较窄分面（如down区域）的箱线图条纹间距明显小于较宽分面（up区域），即使指定了pattern_spacing参数。

解决方案 #

核心问题是geom_boxplot_pattern的图案参数默认基于绘图面板的相对尺寸计算，分面宽度不同时，相同参数会被缩放到不同实际大小。解决方法是将图案间距的单位设置为数据坐标（native），让条纹间距基于箱线图的实际宽度而非面板比例。

修改后的代码：

# generate data
treat1 <- c(rep("bare",10), rep("veg",10), rep("bare",10), rep("veg",10), rep("bare",10), rep("veg",10))
treat2 <- c(rep("wet",10), rep("wet",10), rep("wet",10), rep("wet",10), rep("dry",10), rep("dry",10))
area <- c(rep("down",10), rep("down",10), rep("up",10), rep("up",10), rep("up",10), rep("up",10))
weight <- c(runif(10, 20, 50), runif(10, 1, 15), runif(10, 1, 10), runif(10, 1, 5), runif(10, 1, 10), runif(10, 1, 5))

dat_exp <- tibble("treat1" = treat1, "treat2" = treat2, "area" = area, "mass" = weight) %>%
  mutate(treat2_area = paste(treat2, area, sep = "."),
         treat1_treat2 = paste(treat1, treat2, sep = "."))


# plot it with consistent pattern spacing
ggplot(data = dat_exp, aes(x= treat1_treat2, y = mass, fill = treat2_area, pattern = treat1)) +
  theme_bw() +
  geom_boxplot_pattern(
    position = position_dodge(preserve = "single"), 
    color = "black", 
    pattern_colour = "white", 
    pattern_fill = "white", 
    pattern_angle = 45, 
    pattern_density = 0.3, 
    pattern_spacing = 0.2,  # 根据箱线图宽度调整数值
    pattern_spacing_unit = "native",  # 关键：使用数据坐标单位
    pattern_key_scale_factor = 0.6
  ) +
  scale_pattern_manual(values = c(bare = "stripe", veg = "none")) +
  ggh4x::facet_grid2(. ~ area, space = "free_x", scales = "free", independent = "y", switch = "both")

关键修改说明 #

• 添加pattern_spacing_unit = "native"：将条纹间距的计算基准从面板相对尺寸改为数据坐标，确保不同分面中箱线图的条纹间距一致。
• 调整pattern_spacing数值：因为单位改为数据坐标，原有的小数值会显得过大，需根据实际箱线图宽度调整为合适值（示例中用0.2）。

内容的提问来源于stack exchange，提问作者Em Laskey