Abstract
When building regression models for forecasting, analysts often encounter the problem of multicollinearity or illconditioning in their data sets. In such cases, large variances and covariances can make subset selection and parameter estimation difficult to impossible. In this paper, we suggest several approaches for extending estimation results to forecasting and review theoretical results useful for forecasting with multicollinearity. Several examples are provided.

This publication has 44 references indexed in Scilit: