Hello I am running some k-medoids analysis to see if that algorithm will work better on my dataset compared to the k-means I have been using so far. When running k-means I used to specify the nstart option, equal to a certain number, depending on how many random sets of clusters I wanted. Could anybody tell me if I need to use, or if there is, an equivalent option for the pam() function? pam(x,5) is what I have put together so far. Also, any suggestions for what I might implement there? Thanks Letizia Tomassini, DVM, MPVM, PhD Candidate College of Veterinary Medicine Washington State University [[alternative HTML version deleted]]