First of all: nice forum and lovely to see fellow statisticans and experts here.
I have some general questions concerning cluster analyses. I ran through the net and this forum but couldn't find a straght forward solution.
First of all: I am using both nominal variables and interval variables in my hierarchical analyses. I recoded the nominal vars into binaries like i would do for regression. Does this work ok? I use Z-scores with ward's method SQDIST clustering.
I see a lot of people using the K-means refinement after a hierarchical cluster method. Is this really necessary because i am really happy with ward's initial clusters. When i use k-means on my current data set with Ward's method obtained cluster centroids, my clusters are getting less pretty (as in screwed up)
I hope someone can help,
Thanks in advance!
Advertise on Talk Stats