Question

我可以使用rapply深入查看列表（其结构未知;可能是嵌套的，也可能不是）来提取某些信息？例如，我有一个模型对象列表，其中一个元素也是模型对象列表。我可以一次运行一个函数吗？

df <- data.frame(matrix(rnorm(50), nrow = 10))
names(df) <- c("X","Y1","Y2","Y3","Y4")

modlist <- list(
  m1 = lm(Y1 ~ X, data = df),
  m2 = lm(Y2 ~ X, data = df),
  m3 = list(
    m3.1 = lm(Y3 ~ X, data = df),
    m3.2 = lm(Y4 ~ X, data = df)
  )
)

从每个关系中提取截距：

getcoef <- function(x) coefficients(x)[1]

getcoef(modlist[[1]])
# works fine

rapply(modlist, getcoef, classes = "lm", how = "replace")
# returns unlisted model objects!???

Answer 1

只需重新定义getcoef功能并使用lapply：

> getcoef <- function(x) if(inherits(x, "lm")) coefficients(x)[1] else lapply(x, getcoef)
> str(lapply(modlist, getcoef))
List of 3
 $ m1: Named num -0.143
  ..- attr(*, "names")= chr "(Intercept)"
 $ m2: Named num -0.11
  ..- attr(*, "names")= chr "(Intercept)"
 $ m3:List of 2
  ..$ m3.1: Named num 0.0743
  .. ..- attr(*, "names")= chr "(Intercept)"
  ..$ m3.2: Named num -0.577
  .. ..- attr(*, "names")= chr "(Intercept)"

Answer 2

您可以使用lapply和递归函数：

recSummary <- function(x, ...){
  if("lm" %in% class(x)){
    summary(x, ...)
  }else{
    lapply(x, recSummary, ...)
  }
}

lapply(modlist, recSummary)

rapply似乎在summary

时取消how = replace的结果

更新：提出您的修正问题：

getcoef <- function(x) coefficients(x)[1]
recSummary <- function(x, func = summary, ...){
  if("lm" %in% class(x)){
    func(x, ...)
  }else{
    lapply(x, recSummary, func, ...)
  }
}

lapply(modlist, recSummary, func = getcoef)

$m1
(Intercept) 
0.1913376 

$m2
(Intercept) 
-0.3930023 

$m3
$m3$m3.1
(Intercept) 
0.1184415 

$m3$m3.2
(Intercept) 
0.01367811

Answer 3

这将为您提供所需的信息。我可以用rapply做，但我得到重复，所以这似乎更简单：

> out2 <- unlist(modlist)
> head(names(out2))
[1] "m1.coefficients.(Intercept)" "m1.coefficients.X"          
[3] "m1.residuals.1"              "m1.residuals.2"             
[5] "m1.residuals.3"              "m1.residuals.4"             

> out2[ grepl("coefficients\\.\\(Intercept", names(out2)) ]
$`m1.coefficients.(Intercept)`
[1] -0.1408831

$`m2.coefficients.(Intercept)`
[1] -0.01388696

$`m3.m3.1.coefficients.(Intercept)`
[1] 0.310378

$`m3.m3.2.coefficients.(Intercept)`
[1] -0.4699532

我可以在嵌套的模型对象列表中使用rapply

3 个答案: