Function for accessing regression output

Question

I just wrote this function for personal use only while working on a specific data project. I would be grateful for any feedback!

get.coef1 <- function(x) {
 x <- as.character(x)
 m.names <- paste("fit", x, letters[1:3], sep="")
 models3 <- matrix(numeric(12L), 3, 4)
 for (i in 1:3) {
 models3[i, ] <- summary(get(m.names[i]))$coefficients[2,]
 }
 return(models3)
}

Data to reproduce function above:

x <- rnorm(100)
y <- x^2/4
z <- rnorm(100)+y/100
w <- rnorm(100)

Edit1: And three fitted regressions:

fit3a <- lm(y ~ x)
fit3b <- lm(y ~ x + z)
fit3c <- lm(y ~ x + z + w)

Whose coefficients on x I want to access viaget.coef1(3).

• Phrancis: Welcome to Code Review! I hope you get some great answers.
• flodel: Welcome. Could you show us what input you use when making a call to the function? Also, does the function not assume you have fit{something}{a,b,c} objects in your environment? Can you show how these were built? Thank you.
• snoram: The function assumes I have fit 3 regressions. I just added an example.

Answer 1

Your function has several problematic assumptions for the object names in the environment (get()) and the number of objects (fixed length in the for loop).

In base R it would be more idiomatic and flexible to use some function from the *apply family, which operate on a list of objects, e.g.

do.call(rbind, lapply(list(fit3a, fit3b, fit3c),
 function(x) summary(x)$coefficients["x", ]))
 Estimate Std. Error t value Pr(>|t|)
[1,] 0.01512 0.02927 0.5164 0.6067
[2,] 0.01530 0.02985 0.5127 0.6093
[3,] 0.01732 0.02899 0.5975 0.5516
t(sapply(list(fit3a, fit3b, fit3c),
 function(x) summary(x)$coefficients["x", ]))
 Estimate Std. Error t value Pr(>|t|)
[1,] 0.01512 0.02927 0.5164 0.6067
[2,] 0.01530 0.02985 0.5127 0.6093
[3,] 0.01732 0.02899 0.5975 0.5516

Output of get.coef1(3) (with set.seed(1)):

 [,1] [,2] [,3] [,4]
[1,] 0.01512 0.02927 0.5164 0.6067
[2,] 0.01530 0.02985 0.5127 0.6093
[3,] 0.01732 0.02899 0.5975 0.5516