The data type of the dependent variable in linear regression must be numeric.

• A. Categorical
• B. Numeric
• C. Binary
• D. Ordinal

Answer: (B)

In linear regression, the dependent variable is treated as a numeric quantity that can be added, subtracted, and squared to measure errors. The method estimates a straight-line relationship by minimizing the sum of squared differences between observed numeric values and their predictions. This relies on having a true numeric scale for the outcome, so the residuals and the fit are meaningful.

Categorical outcomes don’t have a natural numeric distance between categories, so forcing them into a linear combination of predictors leads to misinterpretation of coefficients and unreliable predictions. While a binary outcome can be coded as 0/1, using linear regression for it ignores the probability structure and can yield predictions outside the 0–1 range; logistic regression is typically preferred because it models probabilities directly. Ordinal outcomes have an ordered ranking but unequal gaps between levels, which a simple linear model would incorrectly assume; ordinal or multinomial models preserve the appropriate structure.

Therefore, the appropriate data type for the dependent variable in standard linear regression is numeric.