Statistical Comparison of M-Estimators in Multiple Regression Analysis of Nigeria Microfinance Banks Loans and Advances
D. S. A. Wokoma1, F. N. Nwobi2, E. M. Eke3 & A. E. Ukwubuiwe4
1Department of Statistics, Port Harcourt Polytechnic, Rumuola, Rivers State, Nigeria
2Department of Statistics, Imo State University, Owerri, Imo State, 460222, Nigeria
3Department of Statistics, Abia State Polytechnic, P.M.B 7166, Aba, Nigeria
4Department of Statistics, Abia State College of Health Sciences Management Technology, Aba,
Email: dagywoksy@gmail.com,fnnwobi@yahoo.co.uk,,ekeemenaik@yahoo.com,eluojoa@gmail.com
Corresponding Author: D. S. A. Wokoma
ABSTRACT
Residuals of OLS estimation for the classical linear regression model are heavily affected by outliers in the dataset. This leads to unreliable, inefficient and inaccurate parameter estimates that will not yield a robust predictive model of the phenomena. In this paper, we identified a robust estimator in the class of M-estimators that reduced the RMSE of Nigerian Microfinance Loans and advances data by 45.28% relative to the OLS estimate of the RMSE and also recovered the directional effect of Lending interest rate on Loans/advances that the OLS estimator lost as a result of outliers’ effect.
Keywords: Robust, M-estimator, OLS estimator, Outlier, RMSE, Loans and advances, Deposit, Lending interest rate
