Loop panel data using plm

Question

I have a df that looks like this:

Country  Year  X       Y1    Y2    Y3    Y4
ARG      1990  0.4875  1.23  2.51  1.42  4.85
ARG      1991  0.4952  1.66  2.31  1.25  5.02
ARG      1992  0.5120  1.87  2.57  1.92  4.66
ARG      1993  0.5213  1.96  2.41  2.42  5.85

Where X is my independent variable and Ys are the dependent variables I want to loop on.

I tried something like:

for (i in out_start:out_end){
  outcome = colnames(RegTest)[i]
  model <- plm(get(outcome) ~ X, data=RegTest, index=c("Country", "year"), model= "within")
  summary(model)
}

Where I previously set out_start at 4 and end at the end of my columns but I get an error message saying:

Error in uniqval[as.character(effect), , drop = F] : 
  incorrect number of dimensions

I can't see what is wrong since it seems to get correctly the name of the column.

PS: if someone knows also how to list all summaries in one df it would be super helpful.

Answer 1

For solving the issue with the formula You could use the paste() function in conjunction with as.formula().

It is fairly easy to save the results produced by each call of summary() in a list. You could initialize model as a list before the loop and assign the outcome from each iteration of the loop as an entry to model.

The following code chunk demonstrates how to do this using the dataset Fatalities from package AER. I chose miles as the regressor and the loop runs over several measures of traffic fatalities.

library(AER)
library(plm)
data("Fatalities")

out_start <- 17
out_end <- 26

model <- list()

for (i in out_start:out_end){
  outcome <- colnames(Fatalities)[i]
  model[[i-out_start+1]] <- summary(
    plm(as.formula(paste(outcome, " ~ miles", sep = "")), 
        data = Fatalities, 
        index=c("state", "year"), model= "within") 
  )
}
model

Answer 2

Load the following packages

library(tidyverse)
library(plm)
library(broom)

Let's creat your data set: beware, I think the error message depends on "Country" variable which needs more than one value to run the model.

RegTest <- data.frame(rep(c("ARG","SPA"),2),
                      c(1990,1990,1991,1991),
                      c(0.4875,0.4952,0.5120,0.5213),
                      c(1.23,1.66,1.87,1.96),
                      c(2.51,2.31,2.57,2.41),
                      c(1.42,1.25,1.92,2.42),
                      c(4.85,5.02,4.66,5.85))
names(RegTest) <- c("Country","Year","X","Y1","Y2","Y3","Y4")

Pick up the indipendent variables from the original dataset

indipendent_variables <- RegTest[,4:7]

Build a function with your model, setting indipendet_variables like the main argument of the function

plm_function <- function(indipendent_variable) 
    plm(indipendent_variables ~ X, data=RegTest, index=c("Country","Year"), model="within")

Let's use map function from purrr package to run iteratively your function without using for loops

models <- map(indipendent_variables, plm_function)

Inspect model results using tidy function from broom package

tidy(models$Y1)

To get more statistics from model results use augment function from broom package

tidy(augment(models$Y1))

Save augmented results in a dataframe

results_Y1 <- tidy(augment(models$Y1))