Home Pass a vector or string with several elements for the RHS variables to feols

Questions

Pass a vector or string with several elements for the RHS variables to feols

March 15, 2024

I need to run multiple estimations with feols, using sw (stepwise). Since I have many variables for the right-hand side, I would like to pass them as a vector or string with the vector elements. I must me making a beginner’s mistake.

The following code works:

library(fixest)

set.seed(42)
data <- data.frame(
  y = runif(100, min = 1, max = 100),
  x1 = runif(100, min = 1, max = 100),
  x2 = runif(100, min = 1, max = 100)
)

rm <- feols(log(y) ~ sw(log(x1), log(x2)), data=data)
summary(rm)

However, the following does not:

library(fixest)

set.seed(42)
data <- data.frame(
  y = runif(100, min = 1, max = 100),
  x1 = runif(100, min = 1, max = 100),
  x2 = runif(100, min = 1, max = 100)
)

regressors <- c("log(x1)", "log(x2)")
regressors <- paste(regressors, collapse = ", ")

rm <- feols(log(y) ~ sw(regressors), data=data)
summary(rm)

How can I pass right-hand-side variables as a vector or string to feols?

Here is the error message:

Error in feols(log(y) ~ sw(regressors), data = data) : 
  The variable 'regressors' is in the RHS of the formula but not in the data set.

>Solution :

regressors <- c("log(x1)", "log(x2)")
form <- reformulate(sprintf('sw(%s)', toString(regressors)), 'log(y)')
fixest::feols(form, data)

Standard-errors: IID 
Expl. vars.: log(x1)
             Estimate Std. Error  t value   Pr(>|t|)    
(Intercept)  4.268400   0.454189  9.39785 2.4549e-15 ***
log(x1)     -0.168033   0.119207 -1.40959 1.6183e-01    
---
Expl. vars.: log(x2)
            Estimate Std. Error t value  Pr(>|t|)    
(Intercept) 3.031253   0.426176 7.11268 1.878e-10 ***
log(x2)     0.175165   0.117879 1.48597 1.405e-01