The K-Means button is found on the toolbar and under Outlier Detection.

• Number of Clusters: The number of clusters to be calculated.
• Maximum iterations: How many times the calculations will be run, more iterations will refine the results at the cost of longer processing times.
• Determine Number of PCs Using Variance Explained: Can be used to determine the number of PCs (i.e. dimensionality) of the K-Means test using variance explained instead of number of PCs
• Number of PCs (1-25): The number of principal components representing the workspace.
• Scale PC Scores to Variance Explained: Normalizes the scale on the workspace scores using the variance explained.
• Switch Display Style to by Cluster: Automatically changes the display style to cluster, after running a K-Means Test.
• Use Custom Seed For First Centroid: Allows the selection of a custom seed instead of a randomly generated one, creates consistent results across runs.

A more in depth guide on the uses of K-Means and how to run a K-Means test in Sift can be found here.