I do a multiple regression in which spacing between lines is one of four variables. However, this spacing between lines varies so I have grouped them into five classes designated with numbers 1 through 5. The number 1 represents the case of no lines present; the number 2 represents that the spacing between lines exceeds 20 m; 3 represents a spacing from 16 - 20 m; 4 represents spacing from 11 - 15 m; and 5 represents a spacing of 10 m or less. Can the numbers 1 - 5 be called "dummy variables" or do they have to be called something different?. As I understand it, dummy variables can only have the values 1 and 0.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
First, I don't see why this variable should be binned. Binning generally results in a loss of information. For more on binning problems see Vanderbilt U site . The only problem will be "no lines present" but that perhaps ought to be a different variable. If there are nonlinearities in the relationship between spacing and your DV, you can use splines or polynomial terms to deal with it.
Second, if you do decide to bin, then you should probably use bins that are categorical. (Ordinal independent variables are not readily dealt with, as far as I know). Then you would have 4 dummy variables for the 5 levels of your IV.
Another option is to use the bin number as if it was continuous. I can't really recommend this.
answered May 6, 2014 at 0:00