RSCRIPT: call an R script from Stata.


rscript is a Stata command that runs an external R script within Stata. rscript makes debugging and logging easy by displaying R output and error messages in the Stata console and exiting Stata in error when the R script ends in error.


* Determine which version of -rscript- you have installed
which rscript

* Install the most recent version of -rscript-
net install rscript, from("") replace


Calls to rscript must specify the path to the Rscript utility that comes with your R installation. Here are typical locations of the Rscript executable on common operating systems.

* Windows (R version X.Y.Z)
C:/Program Files/R/R-X.Y.Z/bin/Rscript.exe

* Mac and Linux

The rscript option rpath(pathname) specifies the location of the Rscript executable. The default is to call the executable specified by the global macro RSCRIPT_PATH. The following Stata code demonstrates both usages.

* Use the -rpath- option to specify the path to the Rscript executable
rscript using filename.R, rpath("C:/Program Files/R/R-X.Y.Z/bin/Rscript.exe")

* Use global macro RSCRIPT_PATH to specify the path to the Rscript executable
global RSCRIPT_PATH "C:/Program Files/R/R-X.Y.Z/bin/Rscript.exe"
rscript using filename.R

For more details on rscript usage, see the Stata help file included in this package.


This tutorial assumes you have installed the rscript Stata package and have successfully installed R, which is freely available online. You also need to install the following R packages: tidyverse, haven, and estimatr. Install these packages by opening R and executing the following three lines of code:

install.packages('tidyverse', repos='')
install.packages('haven', repos='')
install.packages('estimatr', repos='')

We will write a Stata script that calls an R script, ols_robust.R, and feeds it an input filename and an output filename. The R script will read the input file, estimate an OLS regression with robust standard errors, and write the results to the output file. Here is the code for ols_robust.R:

# Required libraries. You may need to install them first, e.g., install.packages('tidyverse', repos='')

# Parse arguments (if present)
args = commandArgs(trailingOnly = "TRUE")
if (length(args)) {
  arg1 <- args[1]
  arg2 <- args[2]
} else {
  arg1 <- "C:/Program Files/Stata16/ado/base/a/auto.dta"
  arg2 <- "output.csv"

# Estimate OLS model with robust standard errors and display output
my_data <- read_dta(arg1)
ols <- lm_robust(price ~ mpg, data = my_data, se_type = "HC1")

# Outsheet OLS results
write_csv(tidy(ols), arg2)

## EOF

Here is the Stata script:

* Stata: OLS with robust standard errors
sysuse auto, clear
reg price mpg, robust

* R: OLS with robust standard errors
tempfile auto output
save "`auto'", replace
rscript using ols_robust.R, args("`auto'" "`output'")

* Read in the R results
insheet using "`output'", comma clear

The Stata script begins by running the OLS regression in Stata

Stata OLS output

We then save the dataset into a tempfile and call the R script that we wrote. rscript reports that we are calling ols_robust.R and feeding it two arguments, which correspond to the names of the two tempfiles. rscript also reports the output produced by R. We can see here that the point estimates and standard errors are the same as those that were computed by Stata. (Don’t worry about the tidyverse conflicts that are also reported. These namespace conflicts are quite common in R.)

Running rscript

Finally, we read in the results that were outputted from R into Stata and display them. We again have confirmation that that the point estimates and standard errors are the same in both Stata and R.

rscript output

Update History


David Molitor
University of Illinois

Julian Reif
University of Illinois