28. One observation is far from the others and changes the conclusion. How should it be handled?
MediumStatistics & inference
Technical
Show model answer
Check the original record, sample identity, measurement conditions and whether a documented error explains it. An unusual value can be genuine scientific information rather than a reason for automatic exclusion.
Apply justified, preferably predefined criteria, preserve the original data and report the effect of including or excluding it. Do not remove the observation simply because its absence makes the preferred result statistically significant.
What interviewers listen for
- Investigate provenance and error
- Use justified exclusion criteria
- Report sensitivity transparently