Data Transformation

Models may make assumptions on the data they process (see examples below).

Any assumption should be made explicit: if the data coming from a data source
does not meet those assumption, an adaptation/normalization layer should be used
(note: this is different from, and complementary to, the projection/abstraction layer
discussed in another pattern)

Example: Iso-semantic ways of representing ‘quantitative’ data:

1. Stand alone value with units, with unit conversions

2. Relative to a reference value or range (above normal; low-normal-high)

3. Z-score (incorporating mean and SD)

4. Multiple of mean (MOM)

5. Multiple of upper limit of normal (ULN)