如何在表格中呈现 R lm 对象交互系数？

Question

考虑一个带有交互项的 lm fit 对象，例如：

Call:
lm(formula = mpg ~ interaction(gear, am, drop = T) - 1 + cyl, data = mtcars)

Coefficients:
interaction(gear, am, drop = T)3.0  interaction(gear, am, drop = T)4.0  
                            35.478                              34.022  
interaction(gear, am, drop = T)4.1  interaction(gear, am, drop = T)5.1  
                            37.949                              36.946  
                               cyl  
                            -2.594  

是否有一种直接的方法可以在二维表中排列（仅）交互项的系数？例如：

gear    am = 0   am = 1   
3       35.478   NA   
4       34.022   37.949
5       NA       36.946

Answer 1

您可以使用broom 包轻松获取信息。只需根据需要调整表格即可：

model <- lm(formula = mpg ~ interaction(gear, am, drop = T) - 1 + cyl, data = mtcars)

library(broom)
tidy(model)[grep("interaction", tidy(model)$term),]
#                                term estimate std.error statistic      p.value
#1 interaction(gear, am, drop = T)3.0 35.47755  3.533483  10.04039 1.302118e-10
#2 interaction(gear, am, drop = T)4.0 34.02157  2.781402  12.23181 1.599806e-12
#3 interaction(gear, am, drop = T)4.1 37.94942  2.348132  16.16154 2.087857e-15
#4 interaction(gear, am, drop = T)5.1 36.94589  3.096187  11.93271 2.822201e-12

要得到你想要的表，你可以这样做：

dtab <- table(mtcars$gear, mtcars$am)
dtab[which(dtab != 0)] <- tidy(model)[grep("interaction", tidy(model)$term),]$estimate
dtab[which(dtab == 0)] <- NA
  #       0        1
  #3 35.47755         
  #4 34.02157 37.94942
  #5          36.94589