AtlasSession 4L12: Introduction to measurement errors and diagnostic test accuracy studiesp.12–17, 55–56
Random vs systematic error
- Study designs primer
- What is systematic error
- Accuracy vs precision
- Random vs systematic error
- Next: Assessing measurement error
Fig. 1
Rat skin wounds at day 14: collagen area, scaffold-treated vs control, true SD 10 (percentage points)
Fig. 1: 30 wounds per group; arrows run from each wound's true value to its reading. The error is random: fields picked at random in both groups. Control errors add up to about zero, treated errors to about zero.
- Control errors add up to
- about zero
- Treated errors add up to
- about zero
- Fake difference
- none to speak of
Agreement & Error Lab · First look
Level 1 · First look · Lecture 12 · p.12–17, 55–56
Tell errors that cancel out from errors that push every reading the same way.
Tell random from systematic measurement and sampling error by their effect on the average and the variability, and explain the knock-on effects on power and on adjustment for confounders.
Two bathroom scales (the step you are on)
One wobbles a little either way; the other always adds 2 kg.
One bathroom scale wobbles a little either way; another always adds 2 kg. Here each wound’s reading wobbles around its true value: fake difference between the groups, none to speak of.