A normalization error arises when data sources use inconsistent scales, sampling qualities, or aggregation conventions, distorting the distribution of a fused dataset and any models derived from it. section 5.4.1
Before combining sources, check whether:
- Entities were collected on the same basis, through principled selection or comparable observation. section 5.4.1
- Hidden variables, such as offender identity or collection time, are not creating unobserved distortion. section 5.4.1
- Sensors have not changed quantitatively or qualitatively during the collection period. section 5.4.1
Mismatched measurements
Combining agreement scores measured from 1–5 with scores measured from 1–10, or area-average temperatures with point samples, can create a distorted dataset unless the difference is deliberately assessed. section 5.4.1
