stata

Stata Skill — pystata on StataNow 19.5 BE

Run Stata entirely through pystata, the official Python integration that ships with Stata. Configure the session once, then issue every command — and run every .do file — with stata.run(). Data crosses between Python and Stata in memory through pandas, so there is no need to write intermediate .dta files or read .log files unless the user wants them.

The one rule that matters most

Always execute Stata through pystata. Both individual commands and entire .do files go through stata.run(...). Never launch StataBE-64.exe as a subprocess and never run a do-file in batch mode — pystata keeps a single live Stata session in the Python process, gives direct access to data and stored results, and raises real Python exceptions on errors. Running a do-file is just stata.run('do "path/to/file.do"').

Setup

This machine has StataNow 19.5 BE at C:\Program Files\StataNow19, and it is already on PATH. pystata and stata_setup are installed for the system Python (3.14). Basic Edition (BE) is the only licensed edition; "se" and "mp" cannot be initialized.

Configure once per Python process:

import stata_setup
stata_setup.config(r"C:\Program Files\StataNow19", "be")
from pystata import stata

For clean output without the StataCorp splash banner, drive pystata.config directly instead:

import sys
sys.path.insert(0, r"C:\Program Files\StataNow19\utilities")
import pystata
pystata.config.init("be", splash=False)
from pystata import stata

config.init can run only once per process; to start over, launch a fresh Python process.

Bundled helper (optional)

scripts/stata_runner.py removes the boilerplate: it bakes in the path and edition, configures pystata lazily on first use, and wraps command-running, output capture, and data exchange. Reach for it when a script makes several Stata calls.

import os
import sys
sys.path.insert(0, os.path.expanduser("~/.claude/skills/stata/scripts"))
import stata_runner as sr

sr.run("sysuse auto, clear")
log = sr.run("regress price mpg weight, robust", capture=True)
print(log)
print("R-squared:", sr.ereturn()["e(r2)"])