# In cluster analysis, can you use Gower's coefficient of similarity with a k-means clustering method?

I am researching cluster analysis, and I am interested in variables that are both categorical and continuous, for which I have read that a Gower's similarity coefficient is a good proximity measure. I have read that Gower's similarity coefficient is generally not compatible with Ward's method, so I was planning to initially cluster using average linkage, but I was also seeking to compare the cluster structure (for content validity purposes) with another clustering method, specifically the k-means method, using the number of clusters and initial centers obtained in the average linkage method. Is Gower's coefficient of similarity a compatible proximity measure for k-means method?

• Gower's similarity coefficient is generally not compatible with Ward's method You might have also read that it is not comatible with k-means as well. Because k-means and Ward are relatives. Won't you prefer k-medoids in place of k-means? – ttnphns Jul 18 '13 at 8:39
• Thank you-- that's exactly what I needed to know. It looks like k-medoids would be more appropriate. Thanks! – Laura Jul 18 '13 at 11:30