Sample reuse procedures for prediction of the unobserved portion of a partially observed vector

Abstract

Predictive sample reuse methods are applied in several ways to the problem of supplying values for the unobserved portion of a vector when N other similar but fully observed vectors are in hand. One method is to use a mixing matrix to combine two independently calculated vector predictors into a single predictor. The other is to use regression techniques with varying constraints on the regression matrix. The methods are applied to two data sets and for one of them compared with previous prediction analyses.