求助：R语言血清转化率绘图中数据点超出CI问题

我用R语言绘制血清转化率图时，出现数据点超出置信区间（CI）的情况，代码如下：

proportions <- plot_data %>%
  group_by(Age) %>%
  summarize( 
    Prop_pfmsp119 = sum(Sero_pfmsp119 == 1, na.rm = TRUE) / n(),
    Prop_pfama1 = sum(Sero_pfama1== 1, na.rm = TRUE) / n(),
    Prop_pvama1 = sum(Sero_pvama1== 1, na.rm = TRUE) / n(),
    prop_pvmsp119=sum(sero_pvmsp119== 1, na.rm = TRUE)/n()
  )
proportions <- proportions[proportions$Prop_pfmsp119 > 0, ]
proportions <- proportions[proportions$Prop_pfama1 > 0, ]
proportions <- proportions[proportions$Prop_pvama1 > 0, ]
proportions=proportions[proportions$prop_pvmsp119 > 0, ]

# 绘图代码
p1 <- ggplot(mydata, aes(x = Age, y = p)) +
  geom_point(data=proportions,aes(x = Age , y = Prop_pfama1),col="red") +
  # geom_smooth(method="glm",se=T,color="black",method.args = list(family=binomial))+
  #geom_smooth(method = "loess", se = TRUE, color = "black", linewidth = 1) +
  labs(x = "Age (in years)", y = "Probability of Positive") +
  ggtitle("PfAMA1") +
  geom_text(x = 45, y = 0.9, label = "λ = 0.010;95% CI:(0.006-0.014)", color = "black", size = 4) +
  theme_classic() +
  #scale_x_continuous(breaks = seq(0, 100, 20)) +
  ylim(0,1)+
  #geom_ribbon(aes(ymin = l_p, ymax = u_p), fill = "red", linetype = "dashed")
  
  geom_line()+
  geom_line(aes(y=u_p),linetype="dashed")+
  geom_line(aes(y=l_p),linetype="dashed")+
  theme_classic()

绘图后可见红色数据点有部分落在虚线置信区间之外。

问题原因及修正方案 #

1. 数据源不匹配
   绘图时，散点用的是proportions中分组计算的比例，而拟合线和置信区间用的是mydata的p、u_p、l_p，两个数据集的Age分组或数据范围可能不一致，导致点和区间不对应。

2. 修正步骤

   • 统一数据源：将拟合值和置信区间列合并到proportions数据框中，以proportions作为绘图主数据源：

     # 假设已将p、u_p、l_p计算好并添加到proportions中
     p1 <- ggplot(proportions, aes(x = Age)) +
       geom_point(aes(y = Prop_pfama1), col = "red") +
       geom_line(aes(y = p)) +
       geom_line(aes(y = u_p), linetype = "dashed") +
       geom_line(aes(y = l_p), linetype = "dashed") +
       labs(x = "Age (in years)", y = "Probability of Positive") +
       ggtitle("PfAMA1") +
       geom_text(x = 45, y = 0.9, label = "λ = 0.010;95% CI:(0.006-0.014)", color = "black", size = 4) +
       theme_classic() +
       ylim(0,1)
     

   • 直接用geom_smooth拟合并生成置信区间：这种方法无需手动计算区间，更可靠，适合比例数据的二项式拟合：

     p1 <- ggplot(proportions, aes(x = Age, y = Prop_pfama1)) +
       geom_point(col = "red") +
       geom_smooth(method = "glm", se = TRUE, color = "black", method.args = list(family = binomial)) +
       labs(x = "Age (in years)", y = "Probability of Positive") +
       ggtitle("PfAMA1") +
       geom_text(x = 45, y = 0.9, label = "λ = 0.010;95% CI:(0.006-0.014)", color = "black", size = 4) +
       theme_classic() +
       ylim(0,1)
     

   • 检查数据过滤逻辑：当前过滤了所有四个指标均大于0的Age组，若置信区间计算包含这些被删除的组，会导致区间与剩余数据点不匹配，可调整过滤条件或同步置信区间的计算范围。

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