在ggplot中纳入proportion_oac作为协变量绘图时遇错求助

现有如下数据框：

year_group total_AF total_strokes proportion_oac    Sex proportion_per_1000_AF
1   2000-2002      373            10    0.402144772 Female              26.809651
2   2003-2005      557             7    0.359066427 Female              12.567325
3   2006-2008      911             9    0.290889133 Female               9.879254
4   2009-2011     1481            17    0.064145847 Female              11.478731
5   2012-2014     2153            29    0.035764050 Female              13.469577
6   2015-2017     2760            30    0.017028986 Female              10.869565
7   2018-2020     3626            42    0.013513514 Female              11.583012
8   2021-2023     1021             8    0.007835455 Female               7.835455
9   2000-2002      889            18    0.428571429   Male              20.247469
10  2003-2005     1397            20    0.382247674   Male              14.316392
11  2006-2008     1969            33    0.347892331   Male              16.759777
12  2009-2011     2686            31    0.093447506   Male              11.541325
13  2012-2014     3802            42    0.041031036   Male              11.046817
14  2015-2017     4476            46    0.025245755   Male              10.277033
15  2018-2020     5300            64    0.014528302   Male              12.075472
16  2021-2023     1342            12    0.013412817   Male               8.941878

已成功绘制分组柱状图：

ggplot(df_strokeAF, aes(x = year_group, y = proportion_per_1000_AF, fill = Sex)) +
  geom_bar(stat = "identity", position = "dodge", color = "black") +
  labs(x = "Year Group", y = "Proportion per 1000 AF", fill = "Sex") +
  theme_minimal()

尝试添加调整proportion_oac作为协变量的平滑线时，使用以下代码报错：

ggplot(df_strokeAF, aes(x = year_group, y = proportion_per_1000_AF, fill = Sex)) +
  geom_bar(stat = "identity", position = "dodge", color = "black") +
  labs(x = "Year Group", y = "Proportion per 1000 AF", fill = "Sex") +
  geom_smooth(aes(group = Sex, color = Sex), method = "lm", formula = y ~ x * proportion_oac, se = FALSE) +
  theme_minimal()

报错信息：

Warning message:
Computation failed in stat_smooth()
Caused by error:
! object 'proportion_oac' not found

问题原因 #

1. geom_smooth的formula参数中，x仅指代映射到x轴的变量（此处为分类变量year_group），无法直接在公式中引用数据框的其他列（如proportion_oac）。
2. year_group是分类因子，直接作为线性模型自变量不符合建模逻辑，且geom_smooth默认处理方式不支持包含额外协变量的模型拟合。

解决方案 #

要实现调整协变量后的趋势线，需先拟合包含协变量的线性模型，生成调整后的预测值，再将预测值添加到图表中：

1. 拟合模型并生成预测值

# 拟合包含year_group、Sex交互项及proportion_oac协变量的线性模型
model <- lm(proportion_per_1000_AF ~ year_group * Sex + proportion_oac, data = df_strokeAF)

# 生成调整后的预测值，合并到原数据框
df_strokeAF$predicted <- predict(model, newdata = df_strokeAF)

2. 绘制包含趋势线的最终图表

ggplot(df_strokeAF, aes(x = year_group, y = proportion_per_1000_AF, fill = Sex)) +
  geom_bar(stat = "identity", position = "dodge", color = "black") +
  # 添加调整后的趋势线，与柱状图对齐
  geom_line(aes(y = predicted, color = Sex, group = Sex), position = position_dodge(width = 0.9)) +
  labs(x = "Year Group", y = "Proportion per 1000 AF", fill = "Sex", color = "Sex") +
  theme_minimal()

说明 #

• 模型中加入year_group与Sex的交互项，同时纳入协变量proportion_oac，确保预测值是调整该协变量后的结果。
• position_dodge(width = 0.9)用于让趋势线和对应柱状图对齐，宽度与geom_bar默认的dodge宽度匹配。

内容的提问来源于stack exchange，提问作者MaxStudent