PROPOSAL OF NEW TWO-PARAMETER ESTIMATOR FOR GAMMA REGRESSION MODEL WITH CORRELATED REGRESSORS

  • Janet Iyabo Idowu Ladoke Akintola University of Technology, Ogbomoso.
  • Olusoga Akin Fasoranbaku Federal University of Technology, Akure, Ondo State.
  • Kayode Ayinde
Keywords: Gamma regression, Mean squared error, Multicollinearity, Shrinkage parameter, Two-parameter estimator.

Abstract

Multicollinearity among the explanatory variables in the gamma regression model, make the usual maximum likelihood estimator (MLE) for estimating regression parameters in the multiple regression analysis inefficient as the variance of MLE is high and unstable. In recent times, some researchers have proposed estimator based on ridge and Liu biasing parameters to handle the problem of multicollinearity. This paper proposes new two parameter estimators in gamma regression models when there is collinearity among the explanatory variables. Conditions under which the proposed Gamma Modified New Two-Parameter (GMNTP) has better performance are established theoretically and simulation studies were also conducted. Both simulation and real life application results show that, GMNTP estimator with shrinkage parameter  has better performance than the existing estimators in terms of MSE.

 

 

Published
2024-04-01
How to Cite
Idowu, J., Fasoranbaku, O., & Ayinde, K. (2024). PROPOSAL OF NEW TWO-PARAMETER ESTIMATOR FOR GAMMA REGRESSION MODEL WITH CORRELATED REGRESSORS. LAUTECH Journal of Engineering and Technology, 18(1), 64-75. Retrieved from https://laujet.com/index.php/laujet/article/view/626