Which statement best describes discriminant analysis?

• A. It is a clustering method.
• B. It is a prediction model used to classify observations into predefined classes.
• C. It only handles numeric data.
• D. It is a dimensionality reduction technique.

Answer: (B)

Discriminant analysis is a supervised classification method that builds a model from labeled training data to assign new observations to predefined groups. It does this by deriving discriminant functions—linear ones in linear discriminant analysis or quadratic ones in quadratic discriminant analysis—that separate the classes as effectively as possible. When a new observation arrives, its values are plugged into these functions, and the observation is assigned to the class with the highest discriminant score (often tied to the greatest posterior probability). This focus on predicting class labels makes it a prediction model used to classify observations into predefined classes.

It’s different from clustering, which groups data without predefined labels, and it’s not primarily a dimensionality reduction technique, although some discriminant methods can project data into a lower-dimensional space to emphasize class separation. While predictors are usually numeric, categorical variables can be incorporated through appropriate encoding, rather than being a fundamental limitation of the method.