SAV Inspector
Deutsch
Metadata and data are different kinds of change
Renaming a variable, changing its label or editing a value label affects the dictionary. Recoding values or editing a case affects the data matrix. Treat the second category more cautiously because it can alter analyses, bases and audit expectations.
- Variable and file properties
- Value labels and missing-value definitions
- Selected cell values
- Recode, compute and derived variables
- Multiple-response definitions
Stage, review, then write
A change should remain visible before it becomes an output file. Undo and redo help during the working session; an edit recipe records repeatable operations. The writer creates a new SAV or ZSAV copy and the round-trip check immediately reopens it.
The compatibility report compares supported file metadata, dictionary and cases. It reduces accidental corruption but cannot guarantee that every third-party SPSS extension behaves identically in every application.
Use Teams when a change needs approval
Local editing is appropriate for a private working copy. If a colleague must verify exactly what was changed, save a named immutable version in Teams and attach a concise comment. The review then refers to that exact file state.
- Keep the source
Open a copy or rely on the browser workflow, which never overwrites the selected original.
- Make one explainable set of changes
Use the relevant editor or recode preview and inspect the before/after state.
- Run SAV Check
Resolve blockers and review warnings, missing labels and compatibility findings.
- Write and reopen
Create a new SAV/ZSAV, run the round-trip check and name the output clearly.
Frequently asked questions
Can I edit value labels without changing case values?
Yes. Value labels belong to the dictionary and can be changed separately from the underlying numeric or string values.
Does SAV Inspector overwrite the original?
No. It writes a new downloadable file or a new immutable Teams version.
Can the AI assistant edit the file automatically?
No. Structured suggestions are validated and shown as a preview; the user must confirm before a change is staged.
Should every warning block export?
No. Some findings are contextual. Blockers indicate an unsafe or invalid result; warnings should be reviewed and documented rather than automatically hidden.