Optimal Linear Indexes for Some Selection Problems.
Abstract
The problem considered is that of constructing a linear index or linear combination of observable variates so as to maximize the correlation between the index and an unobservable criterion variate, subject to constraints that the correlations between the index and certain other criterion variates be equal to, greater than or equal to, or proportional to specified constants. Under certain circumstances, this problem is mathematically equivalent to that of deriving a selection procedure having a given intensity for choosing individuals or items from a population so as to maximize the post-selection mean of the primary criterion variate subject to restrictions on the post-selection means of the other criterion variates. A problem of this type occurs in selecting cadets for the Air Force Academy, where a battery of tests is administered to each candidate and the results are used in making the selections. Here, the success of a selection procedure is determined by the actual performance of the selected cadets at the Academy, as measured by criterion variates like academic score and military rating. Similar selection problems are commonplace in genetic settings. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1972
- Accession Number
- AD0751257
Entities
People
- David A. Harville
- Thomas E. Reeves
Organizations
- Air Force Research Laboratory