Description
Understanding and using the metadata stored in SPSS .sav files, such as variable labels, value labels, missing-value codes and measurement levels, so survey data move between tools without losing meaning.
A .sav file holds more than a data matrix: it stores variable names, variable labels (the question text), value labels (e.g. 1 = 'Strongly disagree'), user-defined missing values (e.g. 98 = 'Don't know', 99 = 'Refused'), measurement levels, formats and weights. Survey platforms export to .sav precisely because this metadata functions as a codebook. Losing it, for example by exporting to CSV, causes classic errors: 'Don't know' codes averaged as scale points, reversed value orders, and unlabelled variables. Rigour comes from inspecting the dictionary before analysis, using tools such as pyreadstat that read metadata into Python, converting user-missing codes explicitly, and producing a codebook from the file itself.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Siloed
- Stage
- Evaluation, Post-launch
- Data
- Analytical / expert
- Effort
- Hours
Steps
- Import the .sav with metadata (in SPSS, DISPLAY DICTIONARY; in Python, pyreadstat.read_sav returning data and meta).
- Review variable labels, value labels and declared missing codes for every analysis variable.
- Convert user-defined missing codes (e.g. 98, 99, -1) to system missing before computing statistics.
- Verify value label direction for each scale, especially items that will be reverse-coded.
- Generate a codebook from the metadata and store it with the project.
- When writing back to .sav, preserve or add labels so the processed file is self-documenting.
Sources
- SPSS Syntax Editor
- Python integration (SPSS Python plugin)
- pyreadstat for Python ↔ SPSS bridge
- Field, A. (2024) Discovering Statistics Using IBM SPSS Statistics, 6th ed.
- Miller, E. ReadStat library documentation
- The Inquiry Series, C6 — SPSS for UX Researchers — The Right Way
Also known as
- SPSS data file format
- Data dictionary in .sav
Categories
- Research Operations
- Research Design & Rigor