Chernoff faces are a way of representing multivariate data sets with several
(say 4-10) quantitative variables iconically. Each subject is represented by
a cartoon face with oval head, circular arc mouth, oval eyes, nose, and
eyebrows (possibly ears, hair, etc. as well); the shape & position of these
represent the various values. The faces can be put in a grid or used as
scatterplot elements.
The theory is that the human brain detects similarities and differences
between faces, and patterns among them, well for evolutionary reasons. I've
never found them terribly useful myself, but...
"Daedalus", writing his famous humor column in _New_Scientist_ (now in
_Nature_) sugested a similar idea for instrument panels, though the "facial
expressions" were intended to mimic natural emotions. I don't know the order
of priority here! That article is reprinted in "The Inventions of Daedalus"
published about 1985.
A reference:
H. Chernoff, "The Use of Faces to Represent Points in k-dimensional Space
Graphically", Journal of the American Statistical Association, 68, No. 342,
1973, pp. 361-368.
-Robert dawson
