Bayesian Estimation of Different Scale Parameters Using a LINEX Loss Function.

Area of Science:

• Statistics
• Probability Theory
• Mathematical Modeling

Background:

• The Linear Exponential (LINEX) loss function offers asymmetric penalties for estimation errors.
• Traditional estimation methods may not adequately address asymmetric loss scenarios.
• Bayesian inference provides a framework for incorporating prior knowledge into parameter estimation.

Purpose of the Study:

• To develop and evaluate Bayesian estimation techniques using the LINEX loss function.
• To improve parameter estimation for Gamma distributions and Poisson processes.
• To compare LINEX-based methods against Maximum Likelihood Estimation (MLE) and squared error loss.

Main Methods:

• Application of the LINEX loss function in Bayesian estimation for various statistical models.
• Derivation of improved estimators for parameters like mean and median under LINEX loss.
• Sequential prediction of Poisson process intensity using a Bayesian approach and the LINEX loss function.

Main Results:

• Improved estimators for squared mean and the square root of the median were derived.
• Gamma distribution priors yielded Gamma posterior distributions, facilitating LINEX estimation.
• The Approximate Pointwise Optimal (APO) rule was proposed for sequential prediction.
• LINEX estimation demonstrated advantages over squared error loss and MLE in specific scenarios, indicated by lower Mean Square Error (MSE).

Conclusions:

• The LINEX loss function provides a flexible and effective tool for Bayesian estimation, particularly in asymmetric error situations.
• The proposed methods offer enhanced accuracy and robustness for estimating parameters in Gamma and Poisson processes.
• LINEX-based Bayesian approaches are competitive and sometimes superior to traditional methods like MLE.