Practice question · Multiple choice
A dataset contains one value far from the rest. What is wrong with the common advice to "remove the outlier and rerun"?
Hints
- A reaction time of 12 seconds and a reaction time of 0.02 seconds are both outliers. Ask what caused each.
- If you delete it because it changes the p-value, what have you actually done?
Show the answer
A. That it decides on the result rather than the reason
Why
The question is why the point is there, not how far out it sits. A mistyped decimal should go; a genuinely extreme participant is data. Deciding after seeing the effect on the result is the forking-path problem in miniature, which is why exclusion rules belong in the protocol, written before the data arrive.
Practise Data Distributions and Skewness
The app has 6 more questions on this lesson, and keeps your place in the course. Psychology I is free to start.
More questions on Data Distributions and Skewness
- In a right-skewed income distribution the mean sits above the median.
- House prices, hospital stays and reaction times are all right-skewed. What do these variables have in common…
- Sort each real-world variable by its likely skew.
- A charity reports that the mean donation to its campaign was £340. A journalist checks and finds the median…