CODEWITHSUNDEEP
rfunctionevallm
Wade Byron Profe
Wade Byron Profe

Reputation: 176

r function error: object 'y' not found

I'm really struggling to understand why this function is not working. It's just a function to return the coefficients of a simple linear model:

lmfunction <- function(data,y,x,index){
  return(coef(lm(y~x,data=data,subset=index)))
}

Now I create any random dataset:

dataset <- data.frame(x=rnorm(100),y=rnorm(100)*2)

When I try to run the function on dataset I get the following error:

lmfunction(dataset,y,x,1:100)

Error in eval(predvars, data, env) : object 'y' not found

The y variable definitely exists so I'm not understanding this error.

EDIT:

I restarted my r session and the above code worked. However, if I use it on a real dataset I still run in to problems:

data(iris)
lmfunction(iris,Sepal.Length,Petal.Length,1:150)
Error in eval(predvars, data, env) : object 'Sepal.Length' not found

Upvotes: 1

Views: 14024

Answers (2)

AndS.
AndS.

Reputation: 8110

If you want the structure of your function to remain as 4 inputs with no quotes, we can do as follows:

library(dplyr)

lmfunction <- function(data, y, x, index){
  data %>% 
    select(my_y = !!enquo(y), my_x = !!enquo(x)) %>% 
    lm(my_y~my_x, data = ., subset = index) %>% 
    coefficients()
}

lmfunction(iris, Sepal.Length, Petal.Length, 1:150)
#> (Intercept)        my_x 
#>   4.3066034   0.4089223

lmfunction(mtcars, qsec ,mpg, 1:32)
#> (Intercept)        my_x 
#>  15.3547689   0.1241366

What we do is wrap the variables you want in enquo in order to avoid the "object '**' not found" error. Then we simultaneously select the variables and rename them. The reason I renamed was that I found it hard to feed a quosure into lm. Then we run the lm using our newly named variables.

Upvotes: 0

JohnCoene
JohnCoene

Reputation: 2261

Unlike most other functions you probably built you are not passing objects: you are trying to build a formula, hence the error.

lmFunc <- function(data, formula){
    formula <- as.formula(formula)
    lm <- lm(formula, data)
    coef(lm)
}

lmFunc(mtcars, qsec ~ mpg)

Upvotes: 3

Related Questions