numeric and polynomial variables

HikeFan

If a variable is already coded as a dummy, then, after importing the dataset to the design interface, should I set the type as integer or binominal (and then transform from nominal to numeric)? 

Also, if the variable is ordinal should it be set as polynominal or integer? What if it is ordinal but the values are not 'complete' e.g. 1,2,3,4,8?

 

SGolbert

If the variable has been dummy coded, then the type is numeric, by definition of dummy coding.

 

For ordinal variables it doesn't matter if there are "holes". As far as I know they are not explicitely supported by RapidMiner, but the closest thing is a polynominal type: The actual values "1", "2", "100" don't matter. In case that you use it for regression problems, then you will probably have to dummy code it.

 

Best,

Sebastian