Can I fit different regression models using mapply?

Question

I'm trying to see if I can fit multiple regression models using the mapply function on the iris dataset.

I start by defining my regression model:

reg<-function(dependent,independent){lm(eval(paste0(dependent,"~",independent,",data=iris")))}

I then define my vectors of dependent and independent variables:

dependent<-c("Sepal.Length","Sepal.Width")
independent<-c("Sepal.Width","Sepal.Length")

Finally I try to apply the mapply function:

models_reg<-mapply(reg,dependent,independent)

I get an error message, however, claiming that:

Error in parse(text = x, keep.source = FALSE) : 
  <text>:1:25: unexpected ','
1: Sepal.Length~Sepal.Width,

Is it possible to achieve my objective (fitting different regression models using mapply) and in that case, what am I doing wrong?

Answer 1

You don't need to evaluate a string here. You can pass formula as string in lm :

reg<- function(dependent,independent) {
   lm(paste0(dependent,"~",independent),data=iris)
}

Another way to construct the formula is using reformulate :

reg<-function(dependent,independent) {
  lm(reformulate(independent, dependent),data=iris)
}

Now you can call using Map :

Map(reg, dependent, independent)

#$Sepal.Length

#Call:
#lm(formula = reformulate(independent, dependent), data = iris)

#Coefficients:
#(Intercept)  Sepal.Width  
#     6.5262      -0.2234  


#$Sepal.Width

#Call:
#lm(formula = reformulate(independent, dependent), data = iris)

#Coefficients:
# (Intercept)  Sepal.Length  
#     3.41895      -0.06188