## Info

Figure 3 Construction of diagnostic rules for various probability structures. (A) Signal Si has prior probability 0.5, mean 0, and standard deviation 1; S2 has prior probability 0.5, mean 4, and standard deviation 1. (B) S1 has prior probability 0.5, mean 0, and standard deviation 1; S2 has prior probability 0.5, mean 0.1, and standard deviation 1. (C) S1 has prior probability 0.5, mean 0, and standard deviation 1; S2 has prior probability 0.5, mean 1, and standard deviation 1. (D) S1 has prior probability 0.9, mean 0, and standard deviation 1; S2 has prior probability 0.1, mean 4, and standard deviation 1. (See insert for color reproduction of the figure.)

The solution of this equation (z) is the point at which the density functions cross. In cases A, B, and D, the black line is placed at this point. For cases A, B, and C, n = n2 = in case D, n = 9 10 and n2 = 1 10.

This approach is known in statistics as discriminant analysis and in computer science as supervised learning. Sometimes classification is done with logistic regression, which is equivalent when