Normal mixtures is an iterative clustering technique for numerical variables. However, it also predicts the proportion of responses expected within each cluster. Normal mixtures assumes that the joint probability distribution of the measurement columns can be approximated using a mixture of multivariate normal distributions, which represent different clusters. Mean vectors and covariance matrices are estimated for each cluster. See McLachlan and Krishnan (1997) and Section 9.6 in Hand et al. (2001).
Note: The Normal Mixtures algorithm involves iterating through random guesses for the cluster centers. Because of this, results from different runs of the analysis might differ slightly.
If you suspect that you have multivariate outliers, you have two options. You can use an outlier cluster or the Explore Outliers Utility. The outlier cluster option assumes a uniform distribution and is less sensitive to outliers than the standard Normal Mixtures method. The Explore Outliers Utility enables you to explore and handle outliers prior to analysis. See Outlier Cluster and Explore Outliers in Predictive and Specialized Modeling.
Normal Mixtures is one of four platforms that JMP provides for clustering observations. For a comparison of all four methods, see Overview of Platforms for Clustering Observations.