FIND GROUPS
Cluster analysis
Cluster analysis groups observations using similarities in their features. It does not know the “right” groups in advance, so feature choices, measurement scales and validation matter as much as the algorithm.
Groups are a model, not a discovered fact
Different distance measures and algorithms can create different groupings from the same rows. Scale numeric features when their units differ, compare more than one reasonable setup and check whether the resulting groups are useful for the question. Internal scores can help compare solutions, but no single score proves that clusters are real or valuable.
WORKED EXAMPLES
Cluster analysis guides
New guides explain both the mechanics and the limits of each method.
New worked examples will appear here as they are published.