One unusual point can change the picture
Investigate an unusual observation before excluding it.
The short answer
An outlier might reflect a real event, a measurement issue or a change of unit. Inspect the source and definition. If you present a result with the observation excluded, also give the inclusion rule and show how the result differs when it is retained.
A worked example
Consider [10, 11, 50, 12]. The mean is 20.75, while the median is 11.5. The value 50 has a large effect on the mean and on the vertical range of the chart. This does not establish that it is wrong.
Your turn
Calculate the mean of the other three values: 11. Compare it with 20.75. Explain why a different summary is not permission to silently delete the observation.
What this does not tell you
An illustrative calculation does not explain causes or predict future results. Check the dates, units and missing observations before using the method elsewhere.
Related term
OutlierSources & review
Original Charts Haven example. All stated numbers are illustrative and allow you to repeat the calculation.
Next lesson: Look beyond the window