Stata was initially developed by Computing Resource Center in California and the first version was released in 1985.[6] In 1993, the company moved to College Station, Texas and was renamed Stata Corporation, now known as StataCorp.[1] A major release in 2003 included a new graphics system and dialog boxes for all commands.[6] Since then, a new version has been released once every two years.[7] The current version is Stata 18, released in April 2023.[8]
Technical overview and terminology
User interface
From its creation, Stata has always employed an integrated command-line interface. Starting with version 8.0, Stata has included a graphical user interface which uses menus and dialog boxes to give access to many built-in commands. The dataset can be viewed or edited in spreadsheet format. From version 11 on, other commands can be executed while the data browser or editor is opened.
Data structure and storage
Until the release of version 16,[9] Stata could only open a single dataset at any one time. Stata allows for flexibility with assigning data types to data. Its compress command automatically reassigns data to data types that take up less memory without loss of information. Stata utilizes integer storage types which occupy only one or two bytes rather than four, and single-precision (4 bytes) rather than double-precision (8 bytes) is the default for floating-point numbers.
Stata's data format is always tabular in format. Stata refers to the columns of tabular data as variables.
Data format compatibility
Stata can import data in a variety of formats. This includes ASCII data formats (such as CSV or databank formats) and spreadsheet formats (including various Excel formats).
Stata's proprietary file formats have changed over time, although not every Stata release includes a new dataset format. Every version of Stata can read all older dataset formats, and can write both the current and most recent previous dataset format, using the saveold command.[10] Thus, the current Stata release can always open datasets that were created with older versions, but older versions cannot read newer format datasets.
Stata can read and write SAS XPORT format datasets natively, using the fdause and fdasave commands.
Some other econometric applications, including gretl, can directly import Stata file formats.
History
Origins
The development of Stata began in 1984, initially by William (Bill) Gould and later by Sean Becketti. The software was originally intended to compete with statistical programs for personal computers such as SYSTAT and MicroTSP.[6] Stata was written, then as now, in the C programming language, initially for PCs running the DOS operating system. The first version was released in 1985 with 44 commands.[6]
Commands in Stata 1.0 and Stata 1.1
append
dir
infile
plot
spool
beep
do
input
query
summarize
by
drop
label
regress
tabulate
capture
erase
list
rename
test
confirm
exit
macro
replace
type
convert
expand
merge
run
use
correlate
format
modify
save
count
generate
more
set
describe
help
outfile
sort
Development
There have been 18 major releases of Stata between 1985 and 2024, and additional code and documentation updates between major releases.[7] In its early years, extra sets of Stata programs were sometimes sold as "kits" or distributed as Support Disks. With the release of Stata 6 in 1999, updates began to be delivered to users via the web.[6] The initial release of Stata was for the DOS operating system. Since then, versions of Stata have been released for systems running Unix variants like Linux distributions, Windows, and MacOS.[6] All Stata files are platform-independent.
The program command was implemented in Stata 1.2, giving users the ability to add their own commands.[6][13] ado-files followed in Stata 2.1, allowing a user-written program to be automatically loaded into memory. Many user-written ado-files are submitted to the Statistical Software Components Archive hosted by Boston College. StataCorp added an ssc command to allow community-contributed programs to be added directly within Stata.[14] More recent editions of Stata allow users to call Python scripts using commands, as well as allowing Python IDEs like Jupyter Notebooks to import Stata commands.[15] Although Stata does not support R natively, there are user-written extensions to use R scripts in Stata.[16]
User community
A number of important developments were initiated by Stata's active user community.[6] The Stata Technical Bulletin, which often contains user-created commands, was introduced in 1991 and issued six times a year. It was relaunched in 2001 as the peer-reviewed Stata Journal, a quarterly publication containing descriptions of community-contributed commands and tips for the effective use of Stata. In 1994, a listserv began as a hub for users to collaboratively solve coding and technical issues; in 2014, it was converted into a web forum. In 1995, Statacorp began organizing user and developer conferences that meet annually. Only the annual Stata Conference held in the United States is hosted by StataCorp. Other user group meetings are held annually in the United States (the Stata Conference), the UK, Germany, and Italy, and less frequently in several other countries. Local Stata distributors host User Group meetings in their own countries.
Table: Releases and Development of Stata
Version
Release date
Select new or enhanced features
1.0
January 1985
Initial release
Forty-four commands
1.1
February 1985
Bug fixes
1.2
May 1985
New menu system
Better online help
keep
1.3
August 1985
Stata/Graphics
program
1.4
August 1986
New documentation
Formatted infile
1.5
February 1987
anova
logit, probit
2.0
June 1988
New graphics
String variables
Survival analysis: Cox and Kaplan-Meier
Stepwise regression
2.1
September 1990
Byte variables
Factor analysis
ado-files
reshape
3.0
March 1992
logistic, ologit, oprobit, clogit, mlogit
tobit, cnreg, rreg, qreg, weibull, ereg
epitab
pweights
3.1
August 1993
mvreg, sureg, heckman, nlreg, areg, canon
nbreg
constrained linear regression
ml
codebook
4.0
January 1995
xtreg
glm
5.0
October 1996
xtgee, xtprobit
prais, newey, intreg
survey estimation commands
fracpoly
st extended
6.0
January 1999
web aware
new ml
time-series operators
arima, arch
st rewritten
7.0
December 2000
frailty
xtabond
cluster analysis
nlogit
roc
SMCL
8.0
January 2003
graphics
extended GUI, dialog boxes available for all commands
manova
more survey
more time series (VARs, SVARs)
more GLLAMM internalization
8.1
July 2003
updated ml
8.2
October 2003
graphics changes
9.0
April 2005
mata matrix programming language
survey features
linear mixed models
multinominal probit models
9.1
September 2005
9.2
April 2006
10.0
June 2007
graph editor
logistic and Poisson models with complex, nested error components
10.1
August 2008
11.0
July 2009
factor variables
margins postestimation command
multiple imputation
11.1
June 2010
11.2
March 2011
12.0
July 2011
automatic memory management
structural equation modeling
12.1
January 2012
13.0
June 2013
long strings
treatment effects
13.1
October 2013
14.0
April 2015
unicode support
Bayesian statistical analysis
14.1
October 2015
14.2
September 2016
15.0
June 2017
latent class analysis
PDF and Word documents
color transparency or opacity in graphs
15.1
November 2017
16.0
June 2019
frames (multiple datasets in memory)
lasso regression
automated reporting
updated choice models
16.1
February 2020
17.0
April 2021
updated tables command
bayesian econometrics
18.0
April 2023
Bayesian model averaging
causal mediation analysis
heterogeneous difference-in-differences
Software products
There are four builds of Stata: Stata/MP, Stata/SE, Stata/BE, and Numerics by Stata.[17] Whereas Stata/MP allows for built-in parallel processing of certain commands, Stata/SE and Stata/BE are bottlenecked and limit usage to only one single core.[18] Stata/MP runs certain commands about 2.4 times faster, roughly 60% of theoretical maximum efficiency, when running parallel processes on four CPU cores compared to SE or BE versions.[18] Numerics by Stata allows for web integration of Stata commands.
SE and BE versions differ in the amount of memory datasets may utilize. Though Stata/MP can store 10 to 20 billion observations and up to 120,000 variables, Stata/SE and Stata/BE store up to 2.14 billion observations and handle 32,767 variables and 2,048 variables respectively. The maximum number of independent variables in a model is 65,532 variables in Stata/MP, 10,998 variables in Stata/SE, and 798 variables in Stata/BE.[17]
The pricing and licensing of Stata depends on its intended use: business, government/nonprofit, education, or student. Single user licenses are either renewable annually or perpetual. Other license types include a single license for use by concurrent users, a site license, volume single user for bulk pricing, or a student lab.[19]
Example code
The following set of commands revolve around simple data management.[20]
sysuse auto // Open the included auto datasetbrowse// Browse the dataset (opens the Data Editor window)describe// Describes the dataset and associated variablessummarize// Summary information about numerical variablescodebook make foreign // Summary information about the make (string) and foreign (numeric) variablesbrowse ifmissing(rep78) // Browse only observations with missing data for variable rep78list make ifmissing(rep78) // List makes of the cars with missing data for variable rep78
The next set of commands move onto descriptive statistics.
summarize price, detail // Detailed summary statistics for variable pricetabulate foreign // One-way frequency table for variable foreigntabulate rep78 foreign, row // Two-way frequency table for variables rep78 and foreignsummarize mpg if foreign ==1// Summary information about mpg if the car is foreign (the "==" sign tests for equality)by foreign, sort: summarize mpg // As above, but using the "by" prefix.tabulate foreign, summarize(mpg) // As above, but using the tabulate command.
A simple hypothesis test:
ttest mpg, by(foreign) // T-test for difference in means for domestic vs. foreign cars
Graphing data:
twoway (scatter mpg weight) // Scatter plot showing relationship between mpg and weighttwoway (scatter mpg weight), by(foreign, total) // Three graphs for domestic, foreign, and all cars
Linear regression:
generate wtsq = weight^2// Create a new variable for weight squaredregress mpg weight wtsq foreign, vce(robust) // Linear regression of mpg on weight, wtsq, and foreignpredict mpghat // Create a new variable contained the predicted values of mpgtwoway (scatter mpg weight) (line mpghat weight, sort), by(foreign) // Graph data and fitted line
