Allows Stata operations on Julia DataFrames by exporting it to Stata, running a .do file, and re-importing the result into Julia. Requires a copy of Stata.

Using julia > 1.0:

pkg> add StataCall

The package tries to detect your Stata executable automatically by seaching in the most common file paths (under OSX). If it does not find it, it expects you to pass it in the STATA_BIN environment variable.

ENV["STATA_BIN"] = "C:\\Program Files (x86)\\Stata13\\StataMP-64.exe" # this is my location of the Stata executable
using StataCall, DataFrames
df = DataFrame(myint = Int64.(floor.(100 .* rand(Float64, 10))), myfloat = rand(Float64, 10))
instructions = ["gen newvar1 = myint + myfloat";
"gen long newvar2 = floor(_n/2)";
"bysort newvar2: egen double newvar3 = mean(newvar1)"
]
dfOut = StataCall.stataCall(instructions, df)

The main function is stataCall(). The full form is

stataCall(commands::Array{String,1},
     dfIn::DataFrame, 
     retrieveData::Bool = true, 
     doNotEscapeCharacters::Bool = false,
     quiet::Bool = false
     )

• commands is a vector of String that is the series of statements that you want to pass to Stata.
• dfIn is an (optional) DataFrame that you want Stata to open before starting to execute the commands.
• retrieveData is a Bool that says whether you want to retrieve the data after your last command. If true, it will be returned as a DataFrame.
• doNotEscapeCharacters is a Bool that determines whether the strings in commands should be escaped.
• quiet is a Bool that determines whether the Stata output should be suppressed if everything went well. If the script did not finish, output will be displayed.

The function can also be called without the dfIn argument, in which case it starts with an empty dataset.

Please file bug reports on the issues tracker of the Github repo. Fixes and improvements in the form of pull requests are very welcome.

• Log only the body of the Stata output, not headers and footers
• Handle Inf's in DataFrame
• Support Linux; have better way of finding Stata binary under Windows
• Better sync between Julia's variable types and Stata's
• Allow an interactive mode by creating a REPL in Stata and feeding it the commands from Julia. Might be slow because we can pass data only via the hard disk (except under Windows).