Chapter 3 of 10All chapters
Chapter 3 of 10
Beyond the average
Where the average lies.
Look at the spread
An average summarises and conceals. Skew, bimodality and long tails all produce averages that describe nobody in the dataset.
- Percentiles describe real experience better than means for latency and income.
- Two groups mixed together often masquerade as one odd distribution.
Outliers
Decide whether an extreme value is an error, a rare real event, or the most important row in the table. Removing it quietly is a decision that should be recorded.