Chapter 14 Stata

Resources:

Mac Stata does not have the close button . To quit Stata, use either ⌘+Q or exit in the command window.

help <cmd_name>: Get help for a command in Stata console.

For instance, type help estimates to get help for estimates command in Stata. If there are official documentation for the command, you will see “View complete PDF manual entry” hyperlink at the top of the help page. Click on it to view the complete PDF manual.

Overview of Documentation:

  • [U] User’s Guide: is divided into three sections: Stata basics, Elements of Stata, and Advice.

    Recommended to read.

  • Base Reference Manual: list commands alphabetically.

    Not designed to be read from cover to cover.

  • The PDF documentation may be accessed from within Stata.

    help command_name and then click on the “View complete PDF manual entry” button under the command.

    Or in the menu bar, Help > PDF Documentation to open the complete PDF documentation.

    The pdf documentation uses Acrobat Reader as the viewer. Tip: use finger pinch to zoom in and out. When using the zoom button or cmd +/cmd -, the text jumps around, you lose your original position.

Quick Start

// load the auto dataset
. sysuse auto, clear
(1978 automobile data)

// get summary statistics for price and mpg
. summarize price mpg
    Variable |        Obs        Mean    Std. dev.       Min        Max
-------------+---------------------------------------------------------
       price |         74    6165.257    2949.496       3291      15906
         mpg |         74     21.2973    5.785503         12         41

// scatter plot of price and mpg
. scatter price mpg, name(graph1, replace)
============================================================
GRAPHS DETECTED: 1 graph(s) created
============================================================
  • graph1: path-to-dir/graph1.png
Displaying 1 graph(s) in VS Code webview
Displayed 1 graph(s) in VS Code webview

// regress price on mpg
. reg price mpg
      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(1, 72)        =     20.26
       Model |   139449474         1   139449474   Prob > F        =    0.0000
    Residual |   495615923        72  6883554.48   R-squared       =    0.2196
-------------+----------------------------------   Adj R-squared   =    0.2087
       Total |   635065396        73  8699525.97   Root MSE        =    2623.7
------------------------------------------------------------------------------
       price | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
         mpg |  -238.8943   53.07669    -4.50   0.000    -344.7008   -133.0879
       _cons |   11253.06   1170.813     9.61   0.000     8919.088    13587.03
------------------------------------------------------------------------------

sysuse dataset-name [, clear] loads shipped Stata-format datasets. A few datasets are included with Stata and are stored in the system directories. These datasets are often used in the help files to demonstrate a certain feature.

  • clear specifies that it is okay to replace the data in memory, even though the current data have not been saved to disk.

    Only one dataset can be in memory at a time. If you try to load a dataset when another dataset is already in memory, Stata will return an error unless you specify clear to indicate that you want to replace the data in memory.

// list all datasets included with Stata
. sysuse dir
  auto.dta        bplong.dta      citytemp.dta    lifeexp.dta     nlswide1.dta    surface.dta     uslifeexp2.dta
  auto16.dta      bpwide.dta      citytemp4.dta   network1.dta    pop2000.dta     tsline1.dta     voter.dta
  auto2.dta       cancer.dta      educ99gdp.dta   network1a.dta   sandstone.dta   tsline2.dta     xtline1.dta
  autornd.dta     census.dta      gnp96.dta       nlsw88.dta      sp500.dta       uslifeexp.dta

  • The dot (.) indicates that the current line is a Stata command.

  • > indicates that the command is not yet complete. You will see this when you have a command that spans multiple lines.

Keyboard Shortcuts

Actually not of much use.

Keyboard Shortcut Description
ctrl + R last cmd
ctrl + B next cmd
cmd + shift + D Run a Do file

User interface

Within the Stata interface window, there are five windows: Command, Results, History, Properties, and Variables.

Other windows include: Data Editor, Do-file Editor, Graph Editor, Viewer, Variable Manager.

  • Viewer: Help page window.

While Stata can be command-driven by typing code in the Command window, it can also be used in a point-and-click manner using the menu bar.

While nearly everything in Stata can be done via the menus, you’re better off typing commands into a word processing file and saving them, then copying-and-pasting them into the Stata “Command” window.

Buttons

  • Log: Track and save output from the Results window. Ensures replicability.

  • New Do-file Editor: Organize your history commands in one place, making debugging easier.

    You can use do-files to create a batchlike environment in which you place all the commands you want to perform in a file and then instruct Stata to do that file.

    Ex. You have a do file myjob.do, you can run

    do myjob

    all commands in the do file would be sourced.

ref: Stata’s interface

Run Stata in VS Code

You can run Stata in VS Code using the Stata MCP extension.

Use Stata Language for syntax highlighting.

Q: Why run Stata in VS Code?
A: AI integration, code completion, shortcut to run current line, selected lines, etc.

Q: What is MCP?
A: Model Context Protocol (MCP) server.

MCP is a protocol for communication between an LLM (ChatGPT, Claude, etc.) to interact with external tools, applications, databased, etc. It allows the editor to send commands to the LLM and receive responses. The Stata MCP extension implements the MCP protocol to allow AI tools communicate with Stata.

Without MCP, each AI tool needs its own Stata integration. With MCP, any AI tool that supports MCP can integrate with Stata through the Stata MCP server. The MCP server receives requests, executes the corresponding Stata commands, and returns the results to the AI tool.

Further reading:

User case scenario:

  • You write prompts:

    Open my panel dataset, winsorize all financial variables at the 1% level, run FE regressions, and export an esttab table.

  • Then the following workflow will happen:

    User --> LLM --> MCP Server --> Stata --> Results --> MCP Server --> LLM explanation --> User

Configuration

{
  "stata-vscode.stataEdition": "se",
  "stata-vscode.autoStartServer": false,
  "stata-vscode.mcpServerPort": 7001,
}

  • "stata-vscode.stataEdition": "se" to specify the Stata edition you have. Defaults to mp.

    If you have another version other than mp, must specify it accordingly. Otherwise, Stata MCP will not be able to find your Stata executable and activate Stata in VS Code.

  • stata-vscode.autoStartServer: Automatically start MCP server when extension activates. Defaults to true.

    Set to true for projects using Stata; false for projects not using Stata to avoid unnecessary resource usage.

    OUTPUT panel > “MCP: stata-mcp” to see the server status.

  • stata-vscode.mcpServerPort: Port for the MCP server, defaults to 4000.

    4000 conflicts with the default port for bundle jekyll server.

    So I set stata-mcp default port to 7001 to avoid conflict.
    Need to update the port number in your AI tool configuration accordingly, .vscode/mcp.json.

    "servers": {
  "stata-mcp": {
    "type": "http",
    // port number must match the one specified in `stata-vscode.mcpServerPort`
    "url": "http://localhost:7001/mcp-streamable"
  }
}

    Verify local MCP server is listening on. In terminal, run:

    $curl -s http://localhost:7001/health
{"status":"ok","service":"Stata MCP Server","version":"0.4.1","stata_available":true}%

    Paste the set up prompt below into any MCP-aware assistant — Claude Code, OpenAI Codex, Cursor AI, Copilot Chat, etc.

    Remember to use the correct port number you specified in stata-vscode.mcpServerPort.

    Set up the Stata MCP server for me. Endpoint: http://localhost:7001/mcp-streamable — setup guide: https://github.com/hanlulong/stata-mcp#detailed-configurations — if I already have a stata-mcp entry in my MCP config (e.g. using mcp-proxy), replace it rather than appending. When registration succeeds, tell me to restart the client so the stata_run_selection tool becomes available.

    The assistant reads the guide, detects which client it is, writes the right config (or runs the right CLI command), and tells you to restart.

    stata-mcp will be registered in ~/.codex/config.toml. The stata_run_selection tool becomes visible after the restart — MCP tool lists do not refresh mid-session.

Detailed Configuration

Connect to GitHub Copilot, then Copilot can help you:

  • Write and execute Stata commands
  • Analyze your data
  • Generate visualizations
  • Debug Stata code
  • Create statistical reports

Other AI tools that can be integrated with Stata MCP include:

  • ✅ Claude Code, OpenAI Codex.
  • ❌ NO support for Gemini yet.

stata-mcp Skill, GitHub repo

How-to

  • Stata in interactive mode: OUTPUT > choose “Stata”

  • Run Selection / Current Line: ⇧⌘Enter

  • Show outline: Use RegExp Outline extension and add the following to the “Regexp Outline: Header Rules Each Ext” setting.

    [ {"ext": ".do", "rules": [{"level": 1, "format": "^**# (.+)$", "nameIdx": 1, "detail": "H1"}]} ]

    Don’t need to specify level 2, 3, etc. They will be automatically detected by the number of # in the heading.

    • Level 1: **# Level 1 Heading
    • Level 2: **## Level 2 Heading

Do-file

It is recommended to run do files as a whole. (This is different than R.)

You cannot re-run commands freely in Stata.

For example, if you run a command that creates a variable x, realize you made a mistake, and then fix it, you can’t simply select the command that creates x and run it again because x already exists. You could manually drop the existing version of x, but now you’re doing things in a non-reproducible way. Running the entire do file will eliminate this problem because it reloads the data from disk every time. If you find yourself getting confused by these kinds of issues, run the entire do file rather than a selection.

Do-file Rule of Thumb

  • Your .do file begins with loading a dataset and ends with saving one.

    use ..., clear // begin
/* your code */
save ..., replace // end

  • Never modify the raw data files. Save the results of your data cleaning in a new file.

  • Every data file is created by a script. Convert your interactive data cleaning session to a .do file.

  • No data file is modified by multiple scripts.

  • Intermediate steps are saved in different files (or kept in temporary files).

Keep do files short

Our suggestion is that you keep your do files short enough that when you’re working on one of them you can easily wrap your head around it. You also want to keep do files short so they run as quickly as possible: working on a do file usually requires running it repeatedly, so moving any code that you consider “done” to a different do file will save time.

Project Structure

  • You can have a master do file which loads your small section do files sequentially and all in one.

  • Enumerate your do files.

    Example: 0-master.do, 1-data-clean.do, 2-stylized-facts.do, …

    You can then organize them in sub-do-files: if you have different set of stylized facts, you

    could have: 2.1-stylized-facts-geography.do, 2.2-stylized-facts-count.do etc. . . .

log file

log files put everything that your do file put in the Results window.

Comments

  • // for single line comment; rest-of-line comment; it can be put at any place.

    Commonly used after a command to denote comments on that line.

  • * for single line comments; the comment line must begin with *;

  • /* */ for multiple line comment; enclosed comment;

  • //# or **# add a bookmark

    Multiple level bookmarks: add more # to indicate different levels of bookmarks. E.g., //## or **## for level 2 bookmark, //### or **### for level 3 bookmark, etc.

  • /// line-join indicator

    Command panel does NOT support /// line-join indicator. You can only use /// in do-files.

Note that the // comment indicator and the /// indicator must be preceded by one or more blanks.

See [U] 16.1.2 Comments and blank lines in do-files for more details.

Continuation lines: ///

/// is called the line-join indicator or line continuation marker. It makes long lines more readable.

Everything after /// to the end of the current line is considered a comment. The next line joins with the current line. Therefore, /// allows you to split long lines across multiple lines in the do-file.

Summary of ways to break long lines:

  • You can change the end-of-line delimiter to ; by using #delimit,

    #delimit ; // change the line delimiter to semicolon
summarize weight price displ headroom rep78 length turn gear_ratio
    if substr(company,1,4)=="Ford" |
       substr(company,1,2)=="GM", detail ;
gen byte ford = substr(company,1,4)=="Ford" ;
#delimit cr // change the delimiter back to carriage return
gen byte gm = substr(company,1,2)=="GM"

    Once you declear #delimit ;, all lines must end in ;. Stata treats carriage returns as no different from blanks.

  • you can comment out the line break by using /* */ comment delimiters, or

  • you can use the /// line-join indicator.

Example

replace final_result = ///
    sqrt(first_side^2 + second_side^2) ///
    if type == "rectangle"

equivalently, you can use /* */ to break long lines:

replace final_result = /*
    */ sqrt(first_side^2 + second_side^2) /*
    */ if type == "rectangle"

N.B. There’s NO line continuation marker (///) in the command window.

In the command window, the enter key sends what has been written on the line to Stata. There is no way to continue a long command on a second line, without sending the first (incomplete) line to Stata.

You can add comments after ///.

args a /// input parameter for a
     b /// input parameter for b
     c // input parameter for c

is equivalent to

args a b c

cd "directory_name" change working directory.

  • don’t need quotation if there is no space
  • need quotation if the directory has spaces

pwd displays the path of the current working directory.

exit, clear to quit Stata. If the dataset in memory has changed since the last time it was saved, Stata will refuse to quit unless you specify clear.

Abbreviation rules: Stata allows abbreviations. You can abbreviate commands, variable names, and options.

// full command
. summarize myvar, detail
// use abbr. to achieve the same function
. sum myv, d

As a general rule, command, option, and variable names may be abbreviated to the shortest string of characters that uniquely identifies them.

When you read the Stata manual, it uses underlines to denote the minimal abbreviation for a command or option.

E.g. When you see append, it means you can use ap to denote append. describe means the shortest allowable abbreviation for describe is desc.

If there is no underlining, no abbreviation is allowed.

rename can be abbreviated ren, rena, renam, or it can be spelled out in its entirety.

Open do files in tabs rather than in separate windows: https://www.reddit.com/r/stata/comments/1ivjegr/stata_18_mac_does_not_do_tabs_for_dofile_editor/