Two problems are considered: (a) Selecting variables which discriminate well between two or more populations. (6) Estimating log‐likelihood ratios for a particular subject who must be allocated to one or other of the populations. The data on each subject are assumed to be discrete. It is suggested that the use of error rates as a measure of discrimination is unsatisfactory in this situation and an alternative measure is proposed.
No takes yet. Share an insight, caveat, or question.
Michael Hills (1967) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: