SAV Inspector
Deutsch
Three questions for every categorical variable
First ask which values are observed. Then compare them with declared value labels. Finally check whether refusal, not applicable or no answer codes are declared as missing rather than treated as valid responses.
- Observed value without a label
- Declared label never observed
- Same code with inconsistent meaning across variables
- Missing code counted as a valid category
- String spaces, casing or truncation creating duplicate categories
Reuse codeframes, but verify exceptions
Long questionnaires often reuse a small number of label sets across many variables. Grouping identical codeframes makes review faster and helps configure downstream software. A single exceptional item can still differ, so do not apply a shared correction solely because labels look similar.
SAV Inspector can copy codeframes in practical formats and point from a quality finding to the affected variable.
Missing labels are not safely guessed by AI
A variable name such as `satissleep` may suggest a concept but does not prove whether code 1 means good, poor or not applicable. AI can help navigate or prepare a reversible proposal only when the intended mapping is supported by context. Unverifiable labels require a questionnaire, codebook or human owner.
- List observed codes
Scan the complete dataset rather than only the currently filtered rows.
- Compare the dictionary
Identify observed values without labels and inspect missing-value declarations.
- Find shared label sets
Use repeated codeframes to review batteries efficiently, while keeping exceptions visible.
- Document corrections
Use the questionnaire or data owner as evidence, then write and verify a new file.
Frequently asked questions
What is an SPSS value label?
It maps a stored code such as 1 to a human-readable meaning such as “very satisfied”. The case still stores the code.
What is a user-defined missing value?
It is a real stored code that SPSS should exclude from ordinary valid analyses, for example a code representing refusal or not applicable.
Can an unlabelled code still be valid?
Yes. It may be intentional, but it must be checked against the questionnaire or codebook rather than assumed to be an error.
Should unused labels be deleted?
Not automatically. A declared category may be valid even if it does not occur in one wave or subset.