R中调用draw函数可视化bam模型报错：无匹配继承方法

问题描述 #

运行bam模型后调用gratia包的draw()函数时反复报错，已加载gratia和mgcv包，尝试简化模型复杂度、更换拟合方法均未解决。

代码示例 #

library("gratia")
library("mgcv")

m3 <- bam(occur ~ s(temp) + 
        # sal + 
        s(DO) + 
        sed_depth + 
        # water_depth + 
        
        ave_sav + 
        s(prop_sg) + 
        # prop_ma +
        
        Season + 
        TimeOfDay + 
        
        s(ToD) + 
        s(DoY, bs = 'cc') + 
        CYR.std +
        
        # Spatial variation
        s(Longitude, Latitude, bs = 'tp') + 
        
        s(fSite, bs = "re"),
      data = toad, 
      method = 'fREML', 
      # knots = knots,
      # nthreads = 4, 
      discrete = TRUE, 
      family = binomial(link = "logit"), 
      select = TRUE,
      gamma = 1.5)

draw(m3,
     select = c(1,2),
     scale = "fixed")

报错信息 #

Error in (function (classes, fdef, mtable)  : 
  unable to find an inherited method for function ‘draw’ for signature ‘"bam"’

解决方案 #

• 核心原因：gratia的draw()函数仅支持gam类对象，未针对bam类实现对应方法。
• 方法一：转换bam对象为gam类
  bam与gam模型结构高度兼容，手动修改对象类后即可调用draw()：

  class(m3) <- "gam"
  draw(m3, select = c(1,2), scale = "fixed")
  

• 方法二：使用mgcv原生绘图函数
  直接调用mgcv包的plot.gam()实现可视化，无需转换对象：

  plot(m3, select = c(1,2), scale = "fixed")
  

内容的提问来源于stack exchange，提问作者Nate