A MULTIPLE-DECISION APPROACH TO THE SELECTION OF THE BEST SET OF PREDICTOR VARIATES
Abstract
Some 'indifference zone' multiple-decision selection procedure formulations of prediction problems involving multivariate normal populations are considered. These problems are of two types. Part I considers problems involving k bivariate normal populations, where the goal is to select the 'best' population. In this part the 'goodness' of the prediction is measured in terms of three different parameters -- the population conditional variance, the population correlation coefficient, and the absolute value of the population correlation coefficient. Part II considers the problem of selecting the best set of a preassigned number t variates from a set of k predictor variates for predicting a designated variate, the predictand. The 'best' set of predictor variates is defined to be the set of t variates for which the predictand has the smallest population conditional variance (or equivalently the largest population multiple correlation coefficient). Sample size requirements are obtained using asymptotic distribution theory of the transformed statistics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1969
- Accession Number
- AD0696128
Entities
People
- John S. Ramberg
Organizations
- Cornell University