The Data

No Bricks without Clay.

The nature and provenance of data are everything: meditating over these will or ought to be 80%+ of your job.

This means training yourself to ask all the right questions. For example, if you’re building some Personal Health Assistant that operates upon user health data:

1. Who or what produced the data?
   A healthcare system, clinician, patient, sensor, or model?
2. Why was it recorded?
   Care, billing, self-tracking, research?
3. Under what conditions?
   Understaffed rural clinic or urban center? Wearing a cheap wearable that fails on hot and humid days that your aunty gifted you for Christmas?
4. What was recorded?
   Raw signals? Samples? Free text? Structured data? Missingness?
5. How was it recorded?
   Was the store appropriate?
6. What transformations happened?
   Filtering, averaging, imputation?

Data is not a Given

Critically, The Data is Not Separate from the Task. Your modeling/analysis approach is affected by the answers to each and every question up there.