Validation Decisions

Knowledge check · self-test
Validation Decisions
Review wearable validation decisions.
Review wearable validation decisions.
Checkpoint: Sleep Accuracy Claims {.depth-l0}
You now know:
- The sleep validation example uses **120 subjects**, **1 night** each, and reports **74%** epoch-by-epoch accuracy with **0.58** kappa.
- The **26%** complement of classification accuracy is not a standard deviation and cannot be divided by $\sqrt{7}$ to produce a weekly accuracy.
- Weekly trend performance requires repeated-night reference data, bias and agreement analysis, and checks for correlated errors.
:::
Back to Wearable IoT: Validation and Design Trade-offs: Carry the Same Trade-offs into AR Glasses