The Workflow of Data Analysis Using Stata

The Workflow of Data Analysis Using Stata, by J. Scott Long, is an essential productivity tool for data analysts. Aimed at anyone who analyzes data, this book presents an effective strategy for designing and doing data-analytic projects.


In this book, Long presents lessons gained from his experience with numerous academic publications, as a coauthor of the immensely popular Regression Models for Categorical Dependent Variables Using Stata, and as a coauthor of the SPOST routines, which are downloaded over 20,000 times a year.


A workflow of data analysis is a process for managing all aspects of data analysis. Planning, documenting, and organizing your work; cleaning the data; creating, renaming, and verifying variables; performing and presenting statistical analyses; producing replicable results; and archiving what you have done are all integral parts of your workflow.


Long shows how to design and implement efficient workflows for both one-person projects and team projects. Long guides you toward streamlining your workflow, because a good workflow is essential for replicating your work, and replication is essential for good science.


An efficient workflow reduces the time you spend doing data management and lets you produce datasets that are easier to analyze. When you methodically clean your data and carefully choose names and effective labels for your variables, the time you spend doing statistical and graphical analyses will be more productive and more enjoyable.


After introducing workflows and explaining how a better workflow can make it easier to work with data, Long describes planning, organizing, and documenting your work. He then introduces how to write and debug Stata do-files and how to use local and global macros. Long presents conventions that greatly simplify data analysis—conventions for naming, labeling, documenting, and verifying variables. He also covers cleaning, analyzing, and protecting your data.


While describing effective workflows, Long also introduces the concepts of basic data management using Stata and writing Stata do-files. Using real-world examples, Stata commands, and Stata scripts, Long illustrates effective techniques for managing your data and analyses. If you analyze data, this book is recommended for you.


© Copyright 1996–2023 StataCorp LLC

List of tables
List of figures
List of examples
Preface (PDF)

A word about fonts, files, commands, and examples


1 Introduction
1.1 Replication: The guiding principle for workflow
1.2 Steps in the workflow

1.2.1 Cleaning data
1.2.2 Running analysis
1.2.3 Presenting results
1.2.4 Protecting files

1.3 Tasks within each step

1.3.1 Planning
1.3.2 Organization
1.3.3 Documentation
1.3.4 Execution

1.4 Criteria for choosing a workflow

1.4.1 Accuracy
1.4.2 Efficiency
1.4.3 Simplicity
1.4.4 Standardization
1.4.5 Automation
1.4.6 Usability
1.4.7 Scalability

1.5 Changing your workflow
1.6 How the book is organized


2 Planning, organizing, and documenting
2.1 The cycle of data analysis
2.2 Planning
2.3 Organization

2.3.1 Principles for organization
2.3.2 Organizing files and directories
2.3.3 Creating your directory structure

A directory structure for a small project
A directory structure for a large, one-person project
Directories for collaborative projects
Special-purpose directories
Remembering what directories contain
Planning your directory structure
Naming files
Batch files

2.3.4 Moving into a new directory structure (advanced topic)

Example of moving into a new directory structure

2.4 Documentation

2.4.1 What should you document?
2.4.2 Levels of documentation
2.4.3 Suggestions for writing documentation

Evaluating your documentation

2.4.4 The research log

A sample page from a research log
A template for research logs

2.4.5 Codebooks

A codebook based on the survey instrument

2.4.6 Dataset documentation

2.5 Conclusions


3 Writing and debugging do-files

3.1 Three ways to execute commands

3.1.1 The Command window
3.1.2 Dialog boxes
3.1.3 Do-files

3.2 Writing effective do-files

3.2.1 Making do-files robust

Make do-files self-contained
Use version control
Exclude directory information
Include seeds for random numbers

3.2.2 Making do-files legible

Use lots of comments
Use alignment and indentation
Use short lines
Limit your abbreviations
Be consistent

3.2.3 Templates for do-files

Commands that belong in every do-file
A template for simple do-files
A more complex do-file template

3.3 Debugging do-files

3.3.1 Simple errors and how to fix them

Log file is open
Log file already exists
Incorrect command name
Incorrect variable name
Incorrect option
Missing comma before options

3.3.2 Steps for resolving errors

Step 1: Update Stata and user-written programs
Step 2: Start with a clean slate
Step 3: Try other data
Step 4: Assume everything could be wrong
Step 5: Run the program in steps
Step 6: Exclude parts of the do-file
Step 7: Starting over
Step 8: Sometimes it is not your mistake

3.3.3 Example 1: Debugging a subtle syntax error
3.3.4 Example 2: Debugging unanticipated results
3.3.5 Advanced methods for debugging

3.4 How to get help
3.5 Conclusions


4 Automating your work

4.1 Macros

4.1.1 Local and global macros

Local macros
Global macros
Using double quotes when defining macros
Creating long strings

4.1.2 Specifying groups of variables and nested models
4.1.3 Setting options with locals

4.2 Information returned by Stata commands

Using returned results with local macros

4.3 Loops: foreach and forvalues

The foreach command
The forvalues command

4.3.1 Ways to use loops

Loop example 1: Listing variable and value labels
Loop example 2: Creating interaction variables
Loop example 3: Fitting models with alternative measures of education
Loop example 4: Recoding multiple variables the same way
Loop example 5: Creating a macro that holds accumulated information
Loop example 6: Retrieving information returned by Stata

4.3.2 Counters in loops

Using loops to save results to a matrix

4.3.3 Nested loops
4.3.4 Debugging loops

4.4 The include command

4.4.1 Specifying the analysis sample with an include file
4.4.2 Recoding data using include files
4.4.3 Caution when using include files

4.5 Ado-files

4.5.1 A simple program to change directories
4.5.2 Loading and deleting ado-files
4.5.3 Listing variable names and labels
4.5.4 A general program to change your working directory
4.5.5 Words of caution

4.6 Help files

4.6.1 nmlabel.hlp
4.6.2 help me

4.7 Conclusions


5 Names, notes, and labels
5.1 Posting files
5.2 The dual workflow of data management and statistical analysis
5.3 Names, notes, and labels
5.4 Naming do-files

5.4.1 Naming do-files to re-create datasets
5.4.2 Naming do-files to reproduce statistical analysis
5.4.3 Using master do-files

Master log files

5.4.4 A template for naming do-files

Using subdirectories for complex analysis

5.5 Naming and internally documenting datasets

Never name it final!
5.5.1 One time only and temporary datasets
5.5.2 Datasets for larger projects
5.5.3 Labels and notes for datasets
5.5.4 The datasignature command

A workflow using the datasignature command
Changes datasignature does not detect

5.6 Naming variables

5.6.1 The fundamental principle for creating and naming variables
5.6.2 Systems for naming variables

Sequential naming systems
Source naming systems
Mnemonic naming systems

5.6.3 Planning names
5.6.4 Principles for selecting names

Anticipate looking for variables
Use simple, unambiguous names
Try names before you decide

5.7 Labeling variables

5.7.1 Listing variable labels and other information

Changing the order of variables in your dataset

5.7.2 Syntax for label variable
5.7.3 Principles for variable labels

Beware of truncation
Test labels before you post the file

5.7.4 Temporarily changing variable labels
5.7.5 Creating variable labels that include the variable name

5.8 Adding notes to variables

5.8.1 Commands for working with notes

Listing notes
Removing notes
Searching notes

5.8.2 Using macros and loops with notes

5.9 Value labels

5.9.1 Creating value labels is a two-step process

Step 1: Defining labels
Step 2: Assigning labels
Why a two-step system?
Removing labels

5.9.2 Principles for constructing value labels

1) Keep labels short
2) Include the category number
3) Avoid special characters
4) Keeping track of where labels are used

5.9.3 Cleaning value labels
5.9.4 Consistent value labels for missing values
5.9.5 Using loops when assigning value labels

5.10 Using multiple languages

5.10.1 Using label language for different written languages
5.10.2 Using label language for short and long labels

5.11 A workflow for names and labels

Step 1: Plan the changes
Step 2: Archive, clone, and rename
Step 3: Revise variable labels
Step 4: Revise value labels
Step 5: Verify the changes

5.11.1 Step 1: Check the source data

Step 1a: List the current names and labels
Step 1b: Try the current names and labels

5.11.2 Step 2: Create clones and rename variables

Step 2a: Create clones
Step 2b: Create rename commands
Step 2c: Rename variables

5.11.3 Step 3: Revise variable labels

Step 3a: Create variable-label commands
Step 3b: Revise variable labels

5.11.4 Step 4: Revise value labels

Step 4a: List the current labels
Step 4b: Create label define commands to edit
Step 4c: Revise labels and add them to dataset

5.11.5 Step 5: Check the new names and labels

5.12 Conclusions


6 Cleaning your data

6.1 Importing data

6.1.1 Data formats

ASCII data formats
Binary-data formats

6.1.2 Ways to import data

Stata commands to import data
Using other statistical packages to export data
Using a data conversion program

6.1.3 Verifying data conversion

Converting the ISSP 2002 data from Russia

6.2 Verifying variables

6.2.1 Values review

Values review of data about the scientific career
Values review of data on family values

6.2.2 Substantive review

What does time to degree measure?
Examining high-frequency values
Links among variables
Changes in survey questions

6.2.3 Missing-data review

Comparisons and missing values
Creating indicators of whether cases are missing
Using extended missing values
Verifying and expanding missing-data codes
Using include files

6.2.4 Internal consistency review

Consistency in data on the scientific career

6.2.5 Principles for fixing data inconsistencies

6.3 Creating variables for analysis

6.3.1 Principles for creating new variables

New variables get new names
Verify that new variables are correct
Document new variables
Keep the source variables

6.3.2 Core commands for creating variables

The generate command
The clonevar command
The replace command

6.3.3 Creating variables with missing values
6.3.4 Additional commands for creating variables

The recode command
The egen command
The tabulate, generate() command

6.3.5 Labeling variables created by Stata
6.3.6 Verifying that variables are correct

Checking the code
Listing variables
Plotting continuous variables
Tabulating variables
Constructing variables multiple ways

6.4 Saving datasets

6.4.1 Selecting observations

Deleting cases versus creating selection variables

6.4.2 Dropping variables

Selecting variables for the ISSP 2002 Russian data

6.4.3 Ordering variables
6.4.4 Internal documentation
6.4.5 Compressing variables
6.4.6 Running diagnostics

The codebook, problem command
Checking for unique ID variables

6.4.7 Adding a data signature
6.4.8 Saving the file
6.4.9 After a file is saved

6.5 Extended example of preparing data for analysis

Creating control variables
Creating binary indicators of positive attitudes
Creating four-category scales of positive attitudes

6.6 Merging files

6.6.1 Match-merging

Sorting the ID variable

6.6.2 One-to-one merging

Combining unrelated datasets

6.6.3 Forgetting to match-merge

6.7 Conclusions


7 Analyzing data and presenting results

7.1 Planning and organizing statistical analysis

7.1.1 Planning in the large
7.1.2 Planning in the middle
7.1.3 Planning in the small

7.2 Organizing do-files

7.2.1 Using master do-files
7.2.2 What belongs in your do-file?

7.3 Documentation for statistical analysis

7.3.1 The research log and comments in do-files
7.3.2 Documenting the provenance of results

Captions on graphs

7.4 Analyzing data using automation

7.4.1 Locals to define sets of variables
7.4.2 Loops for repeated analyses

Computing t tests using loops
Loops for alternative model specifications

7.4.3 Matrices to collect and print results

Collecting results of t tests
Saving results from nested regressions
Saving results from different transformations of articles

7.4.4 Creating a graph from a matrix
7.4.5 Include files to load data and select your sample

7.5 Baseline statistics
7.6 Replication

7.6.1 Lost or forgotten files
7.6.2 Software and version control
7.6.3 Unknown seed for random numbers

Bootstrap standard errors
Letting Stata set the seed
Training and confirmation samples

7.6.4 Using a global that is not in your do-file

7.7 Presenting results

7.7.1 Creating tables

Using spreadsheets
Regression tables with esttab

7.7.2 Creating graphs

Colors, black, and white
Font size

7.7.3 Tips for papers and presentations


7.8 A project checklist
7.9 Conclusions


8 Protecting your files
8.1 Levels of protection and types of files
8.2 Causes of data loss and issues in recovering a file
8.3 Murphy’s law and rules for copying files
8.4 A workflow for file protection

Part 1: Mirroring active storage
Part 2: Offline backups

8.5 Archival preservation
8.6 Conclusions


9 Conclusions
A How Stata works

A.1 How Stata works

Stata directories
The working directory

A.2 Working on a network
A.3 Customizing Stata

A.3.1 Fonts and window locations
A.3.2 Commands to change preferences

Options that can be set permanently
Options that need to be set each session


Function keys

A.4 Additional resources



Author index (PDF)

Subject index (PDF)


© Copyright 1996–2023 StataCorp LLC

Author: J. Scott Long
ISBN-13: 978-1-59718-047-4
©Copyright: 2009
e-Book version available

The Workflow of Data Analysis Using Stata, by J. Scott Long, is an essential productivity tool for data analysts. Aimed at anyone who analyzes data, this book presents an effective strategy for designing and doing data-analytic projects.
In this book, Long presents lessons gained from his experience with numerous academic publications, as a coauthor of the immensely popular Regression Models for Categorical Dependent Variables Using Stata, and as a coauthor of the SPOST routines, which are downloaded over 20,000 times a year.