The mode is the value (or values) that occur most often in a dataset. Unlike mean and median, the mode applies meaningfully to categorical data (favourite colour, browser type, blood type) where arithmetic doesn't.
A dataset can be:
- Unimodal — one mode (e.g. {1, 2, 2, 3, 4})
- Bimodal — two modes (e.g. {1, 1, 2, 3, 3})
- Multimodal — more than two modes
- No mode — all values appear equally often (typical for continuous data without rounding)
For continuous distributions, "mode" generalises to the peak of the probability density function. The normal distribution is unimodal at its mean; mixture models can be multimodal, signaling distinct subpopulations.
The mode is the most robust to outliers (a single extreme value never becomes the mode), but it ignores most of the data — it's a poor centrality measure for symmetric numerical data, where the mean or median tells you more.