Using Predicted Explanatory Variables and Their Effects on Variance Estimations Under Weighted Least Squares Regression

by Joel Robert Douglas, James R. Knaub, Jr.

Abstract: In the course of collecting sample survey information pertaining to the electric power industry at the Energy Information Administration, the need to utilized non-observed explanatory variable data to predict for other variables became apparent. This is particularly the case with the relationship between natural gas receipts and consumption at electric generating facilities. This paper quantifies the additional variance of prediction which arises from using predicted regressor values, and explores some of the results as applicable to the natural gas receipts and consumption at electric generating facilities as an example.

Key Words: Prediction, imputation , variance, weighted least squares regression, cutoff sampling, classical ratio estimator, residual, beta parameter

Authors:

Authors:
Joel Robert Douglas, joel.douglas@eia.doe.gov
James R. Knaub, Jr., jknaub@eia.doe.gov

Editor: Richard G. Graf,rgraf@sunstroke.sdsu.edu

READING THE ARTICLE: You can read the article in portable document (.pdf) format (49085 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2153 times since February 2, 2010.


Return to the InterStat Home Page.