## Technote (FAQ)

## Question

Will SPSS packages perform cluster analysis via mixture modeling? If so, do they employ the Expectation Maximization (EM) algorithm?

## Answer

SPSS Statistics and SPSS Modeler do not perform cluster analysis via mixture models but SPSS AMOS can perform this analysis.

Clustering methods in Statistics include Hierarchical Cluster (CLUSTER command), K-Means Cluster (QUICK CLUSTER command), TwoStep Cluster (TWOSTEP CLUSTER command) and K Nearest Neighbour (KNN command, which can assign cases to clusters based on their similarity to training cases, without regard to a dependent variable).

Modeler has Modeling nodes for K-Means, TwoStep, and KNN, but not for hierarchical cluster. Modeler also has the Kohonen node, a neural network clustering method that is also known as the Self-Organizing Map. The Auto Cluster node evaluates several cluster models for a set of variables, using any or all of the K-Means, TwoStep, and Kohonen methods as the user chooses. The Statistics Model node will let you run the SPSS Statistics TwoStep or Nearest Neighbor procedures from within Modeler.

SPSS AMOS, the Structural Equation Modeling (SEM) package, allows mixture modeling that can assign cases to clusters (groups in AMOS terms) without regard to a dependent variable. AMOS uses Bayesian Estimation, rather than the EM algorithm, in performing mixture model analysis. See the video at

http://amosdevelopment.com/video/mixture/iriscluster/MixtureIrisCluster.html