PROCEDURES FOR COMPARING THE CORRELATED OUTPUTS OF TWO AIR FORCE BASES,
Abstract
This study is concerned with statistical procedures for simultaneously comparing the outputs of two functional units; possibly the outputs of one Air Force base with those of another. The outputs at a given base are assumed to be random variables and possibly correlated. Various assumptions are made regarding the nature of the correlations among the outputs at a given base and how these correlations relate to those at another base. The statistical tests developed for measuring these outputs are likelihood ratio procedures. In addition to tests of the hypotheses, power functions for the tests are presented, so that sample sizes necessary to achieve a given power can be predetermined. Four separate models are treated, and an example utilizing Air Defense Command data is presented. Tests are also provided for deciding which model to use in any given situation. It it is found that the kinds of tests described will prove useful repeatedly, they can easily be programmed for computer utilization. Computing test statistics involves only algebraic and matric operations, and at worst, a matrix inversion. The procedures described here should help eliminate a great deal of guesswork sometimes associated with large-scale comparisons of dissimilar objects. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1965
- Accession Number
- AD0629129
Entities
People
- S. James Press
Organizations
- RAND Corporation