To create dummy variables for a categorical variable with k groups, how many dummy variables are typically created?

• A. Create as many new variables as the value calculated in step 1
• B. Choose a baseline (i.e., a group against which all other groups should be compared)
• C. Assign that baseline group values of 0 for all of your dummy variables
• D. Count the number of groups you want to recode and subtract 1

Answer: (D)

In regression, you encode a categorical variable with k groups using k−1 dummy variables to avoid perfect multicollinearity. One group is treated as the baseline (reference) and is not represented by a dummy; instead, all dummy variables are 0 for observations in that baseline group. Each of the remaining k−1 dummies indicates membership in a specific non-baseline group, so the regression coefficients represent differences from the baseline category. For example, with four groups you’d create three dummy variables (one for each non-baseline group); observations in the baseline group have all zeros, while observations in other groups have a 1 in the corresponding dummy and 0s in the others. This setup prevents redundancy and keeps coefficients interpretable as comparisons to the baseline.