A missing value records the absence of a measured value. Treat null readings as a data-quality investigation, not as evidence that the underlying measurement was zero. section 5.2.2
Triage missing values with explicit checks:
- Count null readings and compare them with the total record count. section 5.2.2
- Count zero readings separately because zeros may be genuine measurements or substituted bad readings. section 5.2.2
- Add unexplained patterns to the project backlog; concentrated nulls may indicate sensor failure or operating conditions. section 5.2.2
Temperature-reading survey
A table contains 262,944,000 records, including 360,959 null temperature readings—about 0.13%—and 1,890,030 zero readings. The counts suggest limited overall noise, but concentration by temperature or operating condition still requires investigation. section 5.2.2
select count(*) from temperature_readings >262944000
Start with a total-row count, then compare it with the non-null temperature count and the zero-value count. section 5.2.2
