Concept of Cluster Analysis
Key Steps in Cluster Analysis
0. Choose variables (based on theory and previous research)
1. Measures of distance or similarity between objects (measures of proximity)
◦ Depends on the data type: interval, frequency, binary
◦ Distance: geometric measurement. Similarity: content measurement
◦ Calculation of a proximity matrix
2. Forming clusters
◦ Various algorithms: hierarchical / non-hierarchical, agglomerative / divisive, etc.
3. Instruments / criteria for deciding on the number of clusters
◦ Instruments: Agglomerative schedule, structure diagram, dendrogram, icicle plot
◦ Criteria (not available in SPSS): F-value, information criteria etc.
4. Saving and representing cluster membership
◦ Performed by SPSS
5. Interpreting clusters
◦ Taking into consideration the mean values (possibly the variance) of the cluster elements