08-30-2016 07:26 AM
I am using the Variable clustering node in order to choose some variables as input to clustering node. Should I also choose an unsigned variable i.e. a variable from CLUS 0 or should I totally ignore these variables and only choose one variable form CLUS1,CLUS2,... ?
Thanks a lot!
08-30-2016 09:54 AM
I would tend to ignore them unless you have a really strong reason to include them. They could be constants for example.
It sounds like you are using interactive selection, but if you switch to Best Variables you can see that EM rejects the CLUS0 variables.
Hope this helps.
08-30-2016 05:03 PM
Sure, no problem. While sometimes many more variables are used as inputs there is nothing wrong with clustering a small number. The iris data are often used as an example and it only has 4 inputs.