COMPARISON OF SHRINKAGE METHODS FOR SOLVING MULTICOLLINEARITY PROBLEM IN MULTIPLE REGRESSION
COMPARISON OF SHRINKAGE METHODS FOR SOLVING MULTICOLLINEARITY PROBLEM IN MULTIPLE REGRESSION
No Thumbnail Available
Date
2017-06
Authors
USMAN, MOMOH SANI
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Multicollinearity has been a serious problem in Regression analysis; Ordinary Least Regression (OLS) may result in high variability in the estimates of the regression coefficients in the presence of multicollinearity. This study compared four shrinkage methods for solving multicollinearity problem in multiple regressions using data from Nigeria macroeconomic variables such as Gross Domestic Product (GDP) and several variables i.e. Inflation Rate (INFR), Crude Oil Price (COP), Exchange Rate (EXCHR), Interest Rate (INTR), Money Supply (MSUP), and External Reserve (EXTRS) affecting GDP. To study the performance of Shrinkage Methods, Principal Component Regression (PCR), Ridge Regression (RR), Partial Least Square Regression (PLSR), and Centered Regression (CR), were considered for efficiency through Root Mean Square Error (RMSE) and R-square (R2). The result showed that RMSE is smaller for CR as compared to the other prediction techniques; similarly the largest value of R-Square clearly indicates that the CR model is good model. The second smallest values of RMSE and the highest values R-Square belong to the PLSR model. Overall, we conclude that CR and PLSR were the best models amongst all other competing models, i.e. RR and PCR.
Description
A DISSERTATION SUBMITTED TO THE SCHOOL OF POSTGRADUATE STUDIES, AHMADU BELLO UNIVERSITY, ZARIA, NIGERIA IN PARTIAL FULFILLMENT FOR THE AWARD OF MASTER OF SCIENCE (M.Sc) DEGREE IN STATISTICS DEPARTMENT OF STATISTICS AHMADU BELLO UNIVERSITY, ZARIA, NIGERIA
Keywords
COMPARISON,, SHRINKAGE METHODS,, SOLVING MULTICOLLINEARITY PROBLEM,, MULTIPLE REGRESSION,