I have socio-economic data of 60 cities. I wish to create a composite index, for which I require weights. I am planning to use entropy method for this purpose. However, I wish to cluster these cities first and then calculate socio-economic index of each city within each particular cluster (which are independent of each other). My query is that when shall I apply entropy method for deriving weights: before clustering on the entire dataset, or after clustering, within every cluster (say there are 3 clusters with 20 cities each) independently ? Please guide. Also I would be grateful if experts can suggest some other objective weighting measures which are easy to compute.