I have to do binary logistic regression with the independent categorical variables which have more than two levels.
What is the best way to deal with such variables?
You will want to generate indicator variables for each of the "categories" in your categorical variable. You can elect to use k-1 indicators from this variable and interpret relative to your base case (most software has an easy command for this). Or, you can create them and test each, remembering that you can only keep k-1 variables. In this case, you would interpret relative to all others (i.e., 1 if "", 0 otherwise).
Hope that helps.
Jason
Thank you Jason. It was helpful.
I recoded all variables with 3 or more levels in 3 new variables with 1 and 0 and tested all of them.
Best regards
This can be done in two ways while using binary logistic regression which essentially requires a dichotomous dependent variable. First one is by taking the independent variable as categorical variable by specifying the reference level as per requirement. Second one is creating p-1 dichotomous categorical variables if there are p levels in the categorical independent variable.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS