AtlasSession 3L7: Systematic errorsp.47–53
Missing data
Fig. 1
Mice with alginate implants, MSC-laden vs cell-free: vascularised construct at 4 weeks (mice)
Fig. 1. Two groups of mice, one dot each. MSC-laden: 12 vascularised out of 16. Cell-free: 8 vascularised out of 20. Lit: everyone in each group. Dividing gives 1.88 times; the names for these are risk and risk ratio.
- MSC-laden: vascularised
- 12 out of 16
- Cell-free: vascularised
- 8 out of 20
- Dividing (risk ratio)
- 1.88 times
- Subtracting (risk difference)
- about 35 more in every 100
2×2 Workbench · First look
Level 1 · First look · Lecture 7 · p.47–53
See why dropping people whose results went missing for a reason can mislead.
Recognise bias when included participants' outcomes go missing for outcome-related reasons, and test it by re-analysing under favourable and unfavourable assumptions.
Class average trap (the step you are on)
If the weakest students skip the final exam and are dropped, the class looks better.
If the weakest students skip the final exam and are dropped, the class average rises. First count every mouse, with the 4 that died as not vascularised: MSC-laden 12 out of 16.