Chapter Contents |
Previous |
Next |
The DISCRIM Procedure |
The PRIORS statement specifies the prior probabilities of group membership. To set the prior probabilities equal, use
priors equal;To set the prior probabilities proportional to the sample sizes, use
priors proportional;
For other than equal or proportional priors, specify the prior probability for each level of the classification variable. Each class level can be written as either a SAS name or a quoted string, and it must be followed by an equal sign and a numeric constant between zero and one. A SAS name begins with a letter or an underscore and can contain digits as well. Lowercase character values and data values with leading blanks must be enclosed in quotes. For example, to define prior probabilities for each level of Grade, where Grade's values are A, B, C, and D, the PRIORS statement can be
priors A=0.1 B=0.3 C=0.5 D=0.1;
If Grade's values are 'a', 'b', 'c', and 'd', each class level must be written as a quoted string:
priors 'a'=0.1 'b'=0.3 'c'=0.5 'd'=0.1;
If Grade is numeric, with formatted values of '1', '2', and '3', the PRIORS statement can be
priors '1'=0.3 '2'=0.6 '3'=0.1;
The specified class levels must exactly match the formatted values of the CLASS variable. For example, if a CLASS variable C has the format 4.2 and a value 5, the PRIORS statement must specify '5.00', not '5.0' or '5'. If the prior probabilities do not sum to one, these probabilities are scaled proportionally to have the sum equal to one. The default is PRIORS EQUAL.
Chapter Contents |
Previous |
Next |
Top |
Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.