Virata

Type Stata syntax.
Python does the work.

The commands you already know — summarize, regress, svy:, stcox — running on pandas, statsmodels and scipy, on a computer with no Stata licence on it.

Free · Windows, macOS and Ubuntu · works offline once your licence key is activated · nothing from your data ever leaves your computer

The Virata window: a command history and variable list on the left, and
              on the right the output of a survey-weighted mean and a regression,
              laid out as tables.
A survey-weighted mean and a regression, from four typed commands.

It answers in the layout you read fastest

Type the command, get the table — and, if you ask for it, the pandas that produced it. set show python on prints the translation under every result, so you pick up the other language while your actual work gets done.

. regress mpce age i.sector Linear regression Number of obs = 6,000 F statistic = 315.96 Prob > F = 0.0000 R-squared = 0.0953 Root MSE = 1412.299 ------------------------------------------------------------------------------ mpce | Coefficient Std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- sector=Urban | 946.6821 38.61528 24.52 0.000 870.9822 1022.382 age | 5.475203 .9718206 5.63 0.000 3.570085 7.380321 _cons | 3143.552 50.91114 61.75 0.000 3043.748 3243.356 ------------------------------------------------------------------------------ --- equivalent pandas --- import statsmodels.formula.api as smf model = smf.ols("mpce ~ age + C(sector)", data=df).fit()

What you get

Data management, tables, the regression family, and the parts that are hard to put together yourself.

Survey designs, taken seriously

Tell it the design once with svyset — weights, strata, clusters — and svy: carries it into means, totals, proportions, ratios, regression and cross-tabulations. Ignoring the design makes standard errors too small, and nothing warns you.

Small numbers stay private

set mincell 5 hides table cells built from too few people, and audits its own output to catch cells you could work back out. For anyone publishing health, HR or education figures.

Three languages, one dataset

Stata syntax, ten dplyr verbs with set language r, and Excel's SUMIFS family inside generate. Each keeps its own rules about missing values, on purpose — and the manual says where they differ.

Survival and time series

stset, Kaplan–Meier curves, log-rank tests and Cox regression; tsset, moving averages and seasonal decomposition.

22 chart types

Every number is worked out in Python; the page only draws. Charts save as single HTML files you can email, and each one carries the command and the data that made it.

A manual, not a wiki

124 pages, one per command, in plain English with the technical term in brackets. Every example on every page is run each time the manual is built, so the output you read is the output you get.

A Virata line chart of mpce by year for rural and urban households,
            with a Chart and Table toggle in the corner.
Charts come with their own Table view — the numbers behind the picture are always one click away.

Download

Version 0.3.3. Install it, then get a free licence key — it takes a minute and needs one email address.

Windows

Windows 10 and 11, 64-bit Coming soon  
Why isn’t it here?

Built and tested from the same source as the other two -- the Windows installer just is not packaged yet. Check back soon, or use macOS or Ubuntu today.

macOS

Apple silicon, M1 and later Download for macOS Disk image · 116 MB
“Apple could not verify Virata”?

Expected: Virata is not signed with an Apple Developer ID yet, so macOS cannot check who made it. Nothing is wrong with the file.

  1. Click Done — not Move to Bin.
  2. Open System Settings › Privacy & Security.
  3. Scroll to Security, find “Virata” was blocked…, click Open Anyway and confirm with your password or Touch ID.
  4. Open Virata again and click Open.

Right-click › Open no longer works for this on current macOS. When Virata offers to move itself into Applications, macOS may ask once more for the moved copy — the same steps again. After that it opens normally.

Ubuntu

Ubuntu 22.04 and later, 64-bit Coming soon  
Why isn’t it here?

Built and tested from the same source as the other two -- the Ubuntu package just is not packaged yet. Check back soon, or use macOS today.

A licence key is required, and it is free. Virata asks for one the first time you open it. Whether a key is genuine and current is checked on your own computer, every time. The first time you use a key, Virata briefly reaches the internet once to make sure no other computer is already using it — after that, no more network requests for that key, ever. Get a key.

Where we are plain with you

Statistical software that overstates itself is worse than useless. Here is the state of this one, in the same words we would use to a colleague.

Checked against R — never against Stata

Results are compared, number by number, with R's survey, survival, stats and dplyr packages, and the manual says which ones. Nothing here has been compared with Stata itself, because we do not have a Stata licence to compare against. We would rather say that than let you assume otherwise.

Not everything is built yet

Panel data (xtset, xtreg), ARIMA and Stata's date formats are not in this version, and a few options are accepted that should be refused. Every page of the manual opens by saying what that command can do today and what is still coming, and the known bugs are written down rather than hidden.

Not code-signed

Windows and macOS will warn you that they cannot tell who made this, because certificates cost money we have not spent yet. To confirm a download is exactly the file we published, check it against the published checksums.

Your data stays where it is

Virata does the statistics on your own machine and sends none of your data anywhere — which is the point, for anyone working on health, pay or student records. Signing up for a key stores your email address on our server so we can send it to you; that is the only thing we keep.