Categorical data contains labels assigned to things rather than measured quantities, such as colors, species, textures, product identifiers, or sensor types. Survey it to establish category coverage, distribution, and issues requiring deeper analysis. section 5.2.1 section 5.2.3
Turn a categorical-data check into explicit survey work:
- Enumerate distinct category values so the data’s vocabulary and unexpected labels become visible. section 5.2.3
- Count records across categories to identify distributions and decide whether the category set is small enough to inspect directly. section 5.2.3
- Flag values such as UNKNOWN for investigation rather than silently treating them as ordinary categories. section 5.2.3
- Convert important questions into exploratory-data-analysis follow-up items, while stopping the survey before committing to detailed scrutiny. section 5.2.3
Surveying sensor types
For a smart-building temperature_readings table, count distinct sensor_type values before examining their behavior. A result of 7 justifies enumerating A1, A2, Tp, Tn, Tx, Xr, and UNKNOWN, then asking how each type affects readings, ranges, outliers, quality, and measurement granularity. section 5.2.3
SELECT Count(*) FROM (SELECT DISTINCT sensor_type FROM temperature_readings); >7
Inspect the distinct-category count, then pair it with the category listing to decide whether direct enumeration is practical. section 5.2.3
