Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: cluster analysis
Posted:
Jan 17, 2013 3:26 PM


> In fact you are right, using this code I only have the cluster numbers. > Using this cluster method is possible to do what I want, i.e., after > defined clusters (using the code from my first message) apply them to new > data? If yes, can you give me some indication how to proceed please?
Manuel, you might be able to use grpstats to compute the means (centroid) of each cluster on the original data. Then you could use pdist2 and min as in my earlier post to locate the nearest centroid for each new point.
This would be reasonable for your trees based on ward linkage. If you had been using other linkage values, this probably would not be the most appropriate approach.
 Tom



