Question

Is there a function that can extract two or more columns from a coeftest object? This is easy one coeftest object at a time, but can I do the same to a list (other than a for() loop)?

> # meaningless data
> temp <- data.frame(a = rnorm(100, mean = 5), b = rnorm(100, mean = 1),
+                    c = 1:100)
> formulas <- list(a ~ b, a ~ c)
> models <- lapply(formulas, lm, data = temp)
> library(lmtest)
> cts <- lapply(models, coeftest)

> # easy to extract columns one object at a time
> cts[[1]][, 1:2]
              Estimate Std. Error
(Intercept)  5.0314196  0.1333705
b           -0.1039264  0.0987044

> # but more difficult algorithmically
> # either one column
> lapply(cts, "[[", 1)
[[1]]
[1] 5.03142

[[2]]
[1] 5.312007

> # or two
> lapply(cts, "[[", 1:2)
Error in FUN(X[[1L]], ...) : attempt to select more than one element

Maybe the more fundamental question is if there is a way to turn the meat of the coeftest object into a data frame, which would allow me to extract columns singly, then use mapply(). Thanks!

Edit: I would like to end up with a matrices (or data frames) with the first and second columns.

    [[1]]
              Estimate Std. Error
(Intercept)  5.0314196  0.1333705
b           -0.1039264  0.0987044

[[2]]
                Estimate  Std. Error
(Intercept)  5.312007153 0.199485363
c           -0.007378529 0.003429477

Answer 1

[[ is the wrong subset function in this case. Note that when you lapply() over a list, what you are operating on are the components of the list, the bits you would get with list[[i]] where i is the ith component.

As such, you only need the [, 1:2] bit of cts[[1]][, 1:2] in the lapply() call. It is a little bit trickier because of the arguments for [, but easily doable with lapply():

> lapply(cts, `[`, , 1:2)
[[1]]
                Estimate Std. Error
(Intercept)  4.926679544  0.1549482
b           -0.001967657  0.1062437

[[2]]
               Estimate  Std. Error
(Intercept) 4.849041327 0.204342067
c           0.001494454 0.003512972

Note the <space>, before 1:2; this is the equivalent of [ , 1:2].