Chernoff faces Chernoff faces are a way to present a small number of highly multivariate data points to check for outliers. My code allows visualization of eight parameters: head eccentricity, eye size, eye eccentricity, eye spacing, pupil size, eyebrow slant, nose length, mouth size. Along with x and y positions, you can graph ten variables in one scatterplot. The main purpose of this is to visually locate outlying data points: humans are very good at recognizing faces, particularly those of strangers.

The code to produce the faces is the MGgrChernoffFace class in (doc). MG_FACE_DEMO (source) (doc) is an example of using MGgrChernoffFace as a plot symbol.


p>Edward Tufte says about Chernoff faces:

With cartoon faces and even numbers becoming data measures, we would appear to have reached the limit of graphical economy of presentation, imagination, and, let it be admitted, eccentricity.

(The Visual Display of Quantitative Information page 142).