Plot 系数 + 置信区间：根据逻辑条件的点透明度（与 0 重叠）

Question

我想要 plot 点 + 置信区间 (CI)。 点和 CI 应遵循用户定义的调色板，但除此之外，点的填充应根据逻辑条件更改。 更详细地说，我的目标是 plot 多个模型的系数 + 置信区间（不同的响应变量，但同一组解释变量）。 因此，我希望我的系数遵循我建立的调色板：基本上，与相同响应相关的系数应该具有相同的颜色。 除了遵循用户定义的调色板之外，我还希望与 0 重叠的系数是透明的（或者我应该说，具有透明填充，但颜色勾勒出圆周）。 CI 的线条应保持不变（与点颜色相同，实心不透明）。 我已经使用虚拟数据框测试了一些选项，尝试使用填充、透明度（alpha）或将颜色存储在数据框本身中，但到目前为止我一直无法达到预期的结果。 非常感谢任何帮助

library(tidyverse)

# user-defined color palette
response.manual.col.scale <- c("#A6761D", "#FC8D62")

# dummy data frame: coefficients + confidence intervals for two different models (different responses, same predictors)
df <- data.frame(response = c("r1", "r1", "r2", "r2"),
                 explanatory = c("var1", "var2", "var1", "var2"),
                 coef = c(1, 2, 1, 2),
                 lower = c(-1, 1, 0.5, -2),
                 upper = c(4, 3, 1.5, 6)) %>% 
  mutate(
    # variable storing whether the confidence interval overlaps with 0
    coef.zero = if_else((lower/upper) < 0, true = TRUE, false = FALSE),
    # variable storing the colour to be used for fill and colour
    resp.col = rep(c("#A6761D", "#FC8D62"), each = 2),
    resp.fill = if_else(coef.zero == TRUE, true = "black", false = resp.col))

ggplot(data = df,
       aes(x = explanatory,
           y = coef,
           ymin = lower,
           ymax = upper,
           fill = response,
           col = response)) +
  geom_pointrange(position = position_dodge(width = 0.4), size = 2) +
  geom_hline(yintercept = 0) +
  scale_fill_manual(values = response.manual.col.scale) +
  scale_color_manual(values = response.manual.col.scale) +
  theme_bw()

到目前为止，我已经设法获得了这样的 plot：

我希望 plot 看起来像这样：

这些是获得我想要的 plot 的一些不成功的尝试。

ggplot(data = df,
       aes(x = explanatory,
           y = coef,
           ymin = lower,
           ymax = upper,
           fill = if_else(coef.zero == TRUE, true = "transparent", false = response),
           col = response)) +
  geom_pointrange(position = position_dodge(width = 0.4), size = 2) +
  geom_hline(yintercept = 0) +
  # scale_fill_manual(values = response.manual.col.scale) +
  scale_color_manual(values = response.manual.col.scale) +
  theme_bw()


ggplot(data = df,
       aes(x = explanatory,
           y = coef,
           ymin = lower,
           ymax = upper,
           fill = resp.fill,
           col = resp.col)) +
  geom_pointrange(position = position_dodge(width = 0.4), size = 2) +
  geom_hline(yintercept = 0) +
  theme_bw()


ggplot(data = df,
       aes(x = explanatory,
           y = coef,
           ymin = lower,
           ymax = upper,
           fill = response,
           col = response,
           # transparency based on condition
           alpha = if_else(coef.zero == TRUE, true = 0, false = 1))) +
  geom_pointrange(position = position_dodge(width = 0.4), size = 2) +
  geom_hline(yintercept = 0) +
  scale_fill_manual(values = response.manual.col.scale) +
  scale_color_manual(values = response.manual.col.scale) +
  theme_bw() +
  guides(alpha = "none")

提前谢谢了

Answer 1

如果您要努力在数据框中创建 colors，则可以将它们与身份标度一起使用：

df <- data.frame(response = c("r1", "r1", "r2", "r2"),
                 explanatory = c("var1", "var2", "var1", "var2"),
                 coef = c(1, 2, 1, 2),
                 lower = c(-1, 1, 0.5, -2),
                 upper = c(4, 3, 1.5, 6)) %>% 
  mutate(
    # variable storing whether the confidence interval overlaps with 0
    coef.zero = lower/upper < 0,
    # variable storing the colour to be used for fill and colour
    resp.col = rep(c("#A6761D", "#FC8D62"), each = 2),
    resp.fill = if_else(coef.zero == TRUE, true = "white", false = resp.col))

ggplot(data = df,
       aes(x = explanatory,
           y = coef,
           ymin = lower,
           ymax = upper,
           fill = resp.fill,
           col = resp.col)) +
  geom_pointrange(position = position_dodge(width = 0.4), size = 2,
                  shape = 21) +
  geom_hline(yintercept = 0) +
  scale_fill_identity() +
  scale_color_identity() +
  theme_bw()