Website Navigation:
Home >
Journal of Research (Science)

Email Address: jrscience@bzu.edu.pk


<< JRScience Home     Volume 15 No.1   June 2004     << Back to Volume Home

 

ESTIMATION AND ANALYSIS OF REGRESSION COEFFICIENTS WHEN EXPLANATORY VARIABLES ARE CORRELATED

G R Pasha1, Muhammad Akbar Ali Shah2 and Ghosia3

1Department of Statistics, Bahauddin Zakariya University, Multan.

2Department of Statistics, Islamia University, Bahawalpur.

3Institute of Statistics, Punjab University, Lahore.

Abstract: In this paper an unconventional method of the principal components regression is adopted for the solution of multicollinearity and an attempt is made to show that by using this technique, some fairly precise estimates of the co-efficient are obtained. This attractive property of the principal components regression makes it superior to the OLS method while dealing with the multicollinear data. Comparison between the variance of the OLS- estimates and the principal components regression  on income and consumption is made.

Keywords: Latent roots, least square, multicollinearity, principal components, regression.