MIxed type of variables in cluster analaysis

[lionking19063]

Hi,

I am doing a cluster analysis with 10 continuous variables and 3 categorical variables. Instead of converting categorical variables into dummies, I am thinking of creating distance matrix using "PROC DISTANCE".

1) Calculate 3 sets of distance matrix and each set contains the distance between one categorical variable(id category_var1) and 10 continuous variables(var interval(continuous _var1-continuous10) 

2) then merge 3 sets of distance matrix back with the values of 10 continuous variables

3) Standardize them and use standardized variables as the new variables in "PROC CLUSTER" or "PROC FASTCLUS"

Question, Dose the logic make sense to you, particularly step 1 ? Thank you.

[PGStats]

Instead, you could get clusters from continuous_var1-continuous_var10 and test for a relationship between those clusters and your categories with proc freq.

[lionking19063]

You are right. However, I really want to test the effects of categorical variables along with other continuous variables at the same time. Thank you for your response.