A measurement scale tells you which operations preserve meaning. The values may be stored as numbers, but storage format does not decide the scale.
| Scale | What is meaningful | Juniper-style example |
|---|---|---|
| Nominal | Same or different category | Invoice ID; customer segment |
| Ordinal | Category order, not equal gaps | Routine, elevated, high review priority |
| Interval | Equal differences, arbitrary zero | Invoice issue date on a named calendar |
| Ratio | Equal differences and meaningful zero | Calendar days to settlement |
The mean of invoice IDs is nonsense. Coding the three customer segments as 1, 2, and 3 does not create a quantitative distance. The median review priority may identify a middle ordered category, but a difference of one code is not a measured amount.
Settlement days support more operations. Forty days is ten days longer than thirty, and—under the stated time origin—twenty days is twice ten. That does not make every ratio interpretation useful; it only makes the arithmetic meaningful.
Calendar dates provide the business-data contrast. Juniper's packet defines issue date even though the learner extract stores only the resulting duration; generically, the number of days between January 5 and January 15 is meaningful, but a calendar's origin is conventional: one issue date is not meaningfully “twice” another. Dates therefore support differences without turning their stored date codes into ratio-scale amounts.
Match the summary to the question
Nominal variables invite counts and proportions. Ordinal variables can use ordered counts and positional summaries. Interval and ratio variables may support means and standard deviations when their distribution and purpose make those summaries informative. Always preserve the data definition: an allowed operation can still answer the wrong business question.
Put the concept to work
Understand this concept
- Explain which equality, ordering, difference, and ratio comparisons are meaningful for nominal, ordinal, interval, and ratio variables.
Apply this concept
- Classify business-data fields by measurement scale and reject numerical summaries that their coding does not support.
Learning resources
Choose a lesson, try an application, or inspect the sources behind this concept.
Build on these ideas
- Measurement scale — Understand
To apply this concept: Required. Field classification applies the permitted-comparison hierarchy.
- Variable — Understand
To understand this concept: Required. A measurement scale describes the structure of a defined variable.
Lessons
Worked examples and cases
Practice
Common mistaken ideas
Sources
Related concepts
Show 1 more related concepts
Use this idea next
- Distribution — Understand
Required level here: understand. Helpful. The scale limits which order and distance descriptions are meaningful.
- Measurement scale — Apply
Required level here: understand. Required. Field classification applies the permitted-comparison hierarchy.