The advantage of the two-step clustering analysis might be in determining the number of clusters. But again, the choice of the best clustering method depends on your data type and size. Generally, I would take a sample of my data (if data size is too large) and evaluate all of: K-means, Fuzzy C, hierarchical, and two-stage using cluster performance indices (cpi). You can find some articles about cpi in published research.

I think the choice of a clustering method largely depends on the nature of your dataset and what you intend to achieve. I developed and experimented with a two step clustering method for quantising image features (up to and above 100,000) and my aim was to avoid the heavy computation arising from the use of K-means. but if you intend to process 300 samples I don't think you have to worry about computation overhead, therefore k-means is still very useful.

The advantage of the two-step clustering analysis might be in determining the number of clusters. But again, the choice of the best clustering method depends on your data type and size. Generally, I would take a sample of my data (if data size is too large) and evaluate all of: K-means, Fuzzy C, hierarchical, and two-stage using cluster performance indices (cpi). You can find some articles about cpi in published research.

Thank you all. Yes, indeed most of the studies tend to use more than one clustering method in order to get the desired outcome. However, two-step cluster analysis is considered to be a relatively novel method. Thus I am trying to get some feedback from the researchers that have already used this method.