Robust biased estimators for Poisson regression model: Simulation and applications
Adewale F. Lukman, Mohammad Arashi, Vilmos Prokaj
Abstract
Summary The method of maximum likelihood flops when there is linear dependency (multicollinearity) and outlier in the generalized linear models. In this study, we combined the ridge estimator with the transformed M‐estimator (MT) and the conditionally unbiased bounded influence estimator (CE). The two new estimators are called the robust MT estimator and Robust‐CE. A Monte Carlo study revealed that the proposed estimators dominate for the generalized linear models with Poisson response and log link function. The real‐life application results support the simulation outcome.
Topics & Concepts
EstimatorOutlierMulticollinearityMathematicsGeneralized linear modelM-estimatorPoisson distributionStatisticsApplied mathematicsMonte Carlo methodPoisson regressionLinear modelLinear regressionDemographySociologyPopulationAdvanced Statistical Methods and ModelsStatistical Methods and InferenceAdvanced Statistical Process Monitoring