StataCorp Stata

About this release

Stata is the statistics package that economics, epidemiology, political science and public health departments have standardised on, largely because a Stata do file written years ago still runs and still produces the same numbers. Data management, estimation, post estimation testing and publication graphics all live in one program with one syntax, so an analysis never has to be stitched together from three tools.

The command language is the point. Everything done through a menu writes a command to the log, which means a session can be turned into a script that reproduces the entire analysis from raw data to final table. For work that has to survive peer review, that audit trail is worth more than any individual estimator.

The estimator coverage is very wide. Linear and generalised linear models, panel and longitudinal methods, survival analysis, time series, multilevel and mixed models, structural equation modelling, treatment effects, survey weighted estimation and Bayesian methods are all built in with a consistent interface for margins, contrasts and predictions afterwards.

Data handling scales further than most people expect. Frames hold several datasets in memory at once, joins and reshapes are single commands, and the multiprocessor build spreads heavy estimation across every available core. Graphics export at publication quality in vector formats, and tables collect into documents without retyping numbers.

What it does

One language for the whole analysis

Cleaning, estimation, testing and graphing share a single syntax that logs as a reproducible script.

Very broad estimator set

Panel, survival, time series, multilevel, structural equation, survey and Bayesian methods included.

Margins and post estimation

Predictions, contrasts and marginal effects follow any model with the same commands.

Multiple datasets in memory

Frames keep several datasets loaded at once and link between them.

Multiprocessor estimation

The MP build spreads heavy models across all available cores.

Publication graphics

Vector output with full control over every element, exportable straight into a manuscript.

Changed in this version

• Faster estimation on wide panels with many fixed effects.
• Table building commands reworked for cleaner export.
• Improved import of very large delimited files.
• Additional Bayesian model types added.
• Graph editor gains finer control over legend placement.

System requirements

Install order

1. Unpack the archive to a path without accented characters.
2. Run the installer and choose the MP or SE build for the machine.
3. Launch once so the working directory and profile file are created.
4. Set the memory allocation and default working folder before loading a large dataset.

Worth knowing before you start

File details