sjPlot的plot_model无法显示交互项名称问题求助

问题背景 #

运行含多因子交互项的广义线性混合模型后，使用sjPlot包的plot_model()绘制森林图时，Y轴所有交互项均显示为“conditional”，添加auto.labels=T参数无改善；tab_model()输出中交互项标记混用X与冒号:，该问题出现于R、RStudio及所有包更新之后。

模型代码：

modelFINAL<-  glmer(treatment ~ (1|Household_ID)+(1|Case)+(1|Villages)+
                      ProductionSystem*disease+
                      ProductionSystem*costs+  
                      ProductionSystem*education+ 
                      ProductionSystem*age+
                      costs*income+
                      age+
                      income,
                 data = GLM_data, family = binomial(link='logit'), 
                 control=glmerControl(optimizer="bobyqa",optCtrl=list(maxfun=2e5)))

绘图代码：

plot_model(modelFINAL, show.values = TRUE,type='est',sort.est = T,transform=NULL,
 vline.color = "black",title="TITLE")

解决办法 #

1. 手动自定义交互项标签 #

直接通过rename参数为每个交互项指定显示名称，覆盖默认标签：

plot_model(modelFINAL, 
           show.values = TRUE,
           type='est',
           sort.est = TRUE,
           transform=NULL,
           vline.color = "black",
           title="TITLE",
           rename = c(
             "ProductionSystem:disease" = "ProductionSystem × disease",
             "ProductionSystem:costs" = "ProductionSystem × costs",
             "ProductionSystem:education" = "ProductionSystem × education",
             "ProductionSystem:age" = "ProductionSystem × age",
             "costs:income" = "costs × income",
             "age" = "age",
             "income" = "income"
           ))

2. 手动指定模型术语 #

提取模型中的固定效应术语，通过terms参数传递给plot_model()，避免默认标签识别错误：

# 提取模型中的固定效应术语（排除截距项）
fixed_terms <- names(fixef(modelFINAL))[-1]

plot_model(modelFINAL, 
           terms = fixed_terms,
           show.values = TRUE,
           type='est',
           sort.est = TRUE,
           transform=NULL,
           vline.color = "black",
           title="TITLE")

3. 统一tab_model的交互项标记 #

对于tab_model()的标记混用问题，使用transform参数将冒号替换为统一的交互符号：

tab_model(modelFINAL,
          transform = function(x) gsub(":", " × ", x))

也可通过term.labels参数完全自定义所有术语标签：

tab_model(modelFINAL,
          term.labels = c(
            "(Intercept)" = "截距",
            "ProductionSystem" = "ProductionSystem",
            "disease" = "disease",
            "ProductionSystem:disease" = "ProductionSystem × disease",
            "costs" = "costs",
            "ProductionSystem:costs" = "ProductionSystem × costs",
            "education" = "education",
            "ProductionSystem:education" = "ProductionSystem × education",
            "age" = "age",
            "ProductionSystem:age" = "ProductionSystem × age",
            "income" = "income",
            "costs:income" = "costs × income"
          ))

4. 回退至稳定版本 #

若确定为包更新导致的bug，可回退至sjPlot或lme4的旧版本：

# 先安装devtools
install.packages("devtools")
# 回退至sjPlot 2.8.10版本（示例，可根据实际情况调整）
devtools::install_version("sjPlot", version = "2.8.10", repos = "http://cran.us.r-project.org")

内容的提问来源于stack exchange，提问作者Kathrin Loosli