On the Restricted Poisson Ridge Regression Estimator

For modeling count data, the Poisson regression model is widely used in which response variable takes non-negative integer values. However, presence of strong correlation between explanatory variables causes problem multicollinearity. Due to multicollinearity, variance maximum likelihood estimator (MLE) will be inflated causing parameters estimation become unstable. Multicollinearity can tackled by using biased estimators such as ridge order minimize estimated coefficients. An alternative approach specify exact linear restrictions on addition model. In this paper, restricted (RPRRE) suggested handle multicollinearity with parameters. addition, conditions superiority comparison some existing are discussed based mean squared error (MSE) matrix criterion. Moreover, a simulation study and real data application provided illustrate theoretical results. The results indicate that estimator, RPRRE, outperforms other terms scalar (SMSE). Therefore, it recommended use RPRRE for when present.