如何去除ggplot中第二个图例标题？glmer绘图问题排查

问题说明 #

用ggpredict绘制glmer模型的预测结果时，遇到两个问题：

• 图例异常：要么出现两个重复图例，要么图例没有颜色（疑似重复设置颜色编码导致）
• 不清楚图中虚线的来源

用户原始代码 #

plot2d = ggpredict(m2d, terms = c("Verbstellung", "group"))
plot(plot2d, connect_lines = TRUE, colors = c("red", "blue")) +
  aes(linetype = .data[["group"]]) +
  ggeasy::easy_center_title()+
  ggtitle("Accuracy in Test Session \n Acceptable Sentences only") +
  theme(plot.title = element_text(size = 18, face = "bold")) +
  xlab("Verb Position") + ylab("Accuracy in Percent")+ 
  easy_all_text_size(15)+
  easy_all_text_color("black")+ 
  scale_color_manual(name = "Group", values = c("red", "blue"), labels = c("LR", "NLR"))

示例数据 #

group   subj    Verbstellung    answer.corr Set actor2
nlr 38  V1  1   2   typical
nlr 38  V1  0   3   typical
nlr 38  V1  1   1   typical
nlr 38  V2  1   3   typical
nlr 38  V3  0   3   typical
nlr 38  V3  1   1   typical
nlr 38  V3  1   2   typical
nlr 38  V2  1   1   typical
lr  35  V1  0   1   typical
lr  35  V3  1   2   typical
lr  35  V2  1   3   typical
lr  35  V2  0   1   typical
lr  35  V3  0   1   typical
lr  35  V1  1   3   typical
lr  35  V3  1   3   typical
lr  35  V1  1   2   typical

所用glmer模型 #

m2d = glmer(answer.corr ~ group + actor2 + Verbstellung + group:actor2 + group:Verbstellung + (1|subj) + (1|Set), data=data, family="binomial", control = glmerControl(optimizer="nlopt", optCtrl=list(maxfun=1e5), calc.derivs = FALSE))

模型输出 #

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
 Family: binomial  ( logit )
Formula: answer.corr ~ group + actor2 + Verbstellung + group:actor2 +      group:Verbstellung + (1 | subj) + (1 | Set)
   Data: data_akz
      AIC       BIC    logLik  deviance  df.resid 
 4033.361  4101.507 -2006.680  4013.361      6722 
Random effects:
 Groups Name        Std.Dev.
 subj   (Intercept) 1.5788  
 Set    (Intercept) 0.1448  
Number of obs: 6732, groups:  subj, 53; Set, 3
Fixed Effects:
          (Intercept)                 group1                actor21         VerbstellungV2         VerbstellungV3         group1:actor21  group1:VerbstellungV2  group1:VerbstellungV3  
               2.3894                 0.5053                 0.4712                 0.4055                 0.7303                 0.1013                -0.3174                -0.5449  

问题原因 #

1. 图例异常：

   • plot(plot2d, colors = c("red", "blue"))已经自动为group变量绑定了颜色映射
   • 后续手动添加aes(linetype = group)和scale_color_manual，相当于重复定义颜色映射规则，导致ggplot生成冲突的图例，出现重复或无颜色的情况

2. 虚线来源：

   • 图中的虚线是模型预测值的置信区间（Confidence Interval），是ggpredict绘图时默认添加的，用来展示预测结果的不确定性范围

修正后的代码 #

直接基于ggpredict返回的预测数据手动构建ggplot，避免自动绘图的冲突：

library(ggplot2)
library(ggeasy)

# 获取预测数据
plot2d = ggpredict(m2d, terms = c("Verbstellung", "group"))

# 手动绘制ggplot
ggplot(plot2d, aes(x = x, y = predicted, color = group, linetype = group)) +
  # 绘制预测线
  geom_line(linewidth = 1) +
  # 绘制置信区间（虚线，可根据需要删除）
  geom_ribbon(aes(ymin = conf.low, ymax = conf.high, fill = group), alpha = 0.2, linetype = "dashed") +
  # 统一设置颜色、线型和填充色，合并图例
  scale_color_manual(name = "Group", values = c("red", "blue"), labels = c("LR", "NLR")) +
  scale_linetype_manual(name = "Group", values = c("solid", "dashed"), labels = c("LR", "NLR")) +
  scale_fill_manual(name = "Group", values = c("red", "blue"), labels = c("LR", "NLR")) +
  # 标题与标签设置
  ggtitle("Accuracy in Test Session \n Acceptable Sentences only") +
  xlab("Verb Position") + 
  ylab("Accuracy in Percent") +
  # 主题调整
  easy_center_title() +
  theme(plot.title = element_text(size = 18, face = "bold")) +
  easy_all_text_size(15) +
  easy_all_text_color("black")

代码说明 #

• 统一在aes中把color和linetype绑定到group，确保图例自动合并为一个
• 用scale_*_manual系列函数统一设置图例的名称、颜色/线型/填充色，避免映射冲突
• 如果不需要置信区间，直接删除geom_ribbon部分即可
• 手动绘图比ggpredict的plot()方法更灵活，能完全控制所有绘图元素

内容的提问来源于stack exchange，提问作者afiw