Testing Clustering Variables I have two clusters. Those two clusters were obtained using Fuzzy C-Means with 8 variables. I'd like to know which variables have important role in differentiating the two clusters. Can I use t test for each variables to see if the two clusters have different mean in those variables? Do you any better idea what test I should use? 
 A: There's a couple things here. In response to your question directly, that's probably not the best idea unless you're going to correct for the fact that making multiple comparisons. Briefly, since this topic has been covered pretty heavily, if you have a lot of variable pairs and you're testing to see whether there is a difference in means between the variables, if you compute a lot of pairwise t-tests you'll end up rejecting null hypothesis purely by chance (check out the classification of m tests in https://en.wikipedia.org/wiki/Multiple_comparisons_problem this page for a good description). What I would recommend is either doing those t-test but then correcting with some method (hochbergs, FDR, bonferoni).
But aside from that, I'm a little confused at how you're using FCM. FCM typically gives back weights that tell you how much a particular observation belongs to a particular cluster. Are you just saying "whichever cluster has a higher weight for observation x is the cluster that observation x belongs to"?
