# Using linear discriminant analysis to validate the cluster groups resulting from kmeans

I'm currently working on a cluster analysis project and ran kmeans on the data for k=2.

I was reading similar articles on similar experiments, and the investigators used discriminant analysis to verify their cluster groups.

I was wondering how to use linear discriminant analysis is used to verify the cluster groups from kmeans, is LDA done separately and then done compared to the kmeans result or if a result from the kmeans is used as a parameter in the LDA call in R?

I'm sorry if this question is vague, but I have never been exposed to discriminant analysis before.

Also the data gathered is repeated measures, and unbalanced per individual, not sure if that valids some assumptions for LDA.

• I was reading similar articles on similar experiments, and the investigators used discriminant analysis to verify their cluster groups Didn't those articles people tell what they did and why? – ttnphns Mar 18 '13 at 13:18
• No they just mentioned it briefly it wasn't a statistics paper just a biology one – DJ_ Mar 18 '13 at 18:06
• Well, if by "verify" of "validate" you mean to check that there naturally exist 2 rather than 1 or 3 or 4 clusters, use Gap clustering index or similar. The main problem with LDA for such a task is that with 2 clusters you get only 1 discriminant out of all the variables (which can be plenty) and due to that you may loose sufficient amount of information about the separatedness of the clusters. – ttnphns Mar 18 '13 at 18:30
• I did use the gap statistic if that is what you mean – DJ_ Mar 18 '13 at 18:48