Statistical Inference and Prediction for Multiply-Hybrid Censored Data with Applications

Document Type : Original Article

Authors

1 Statistics, The Politics and Economics faculty- Beni-Suef University

2 statistics department, Commerce faculty, Aswan University, Egypt

3 Statistics departement, the Politics and Economics faculty, Beni-Suef University

10.21608/esju.2025.351499.1063

Abstract

This study aims to enhance the prediction of failure times for specific units in lifetime experiments where complete observation of all failure times is impractical, introducing a novel approach to handling multiply-hybrid censored data. Leveraging the power Lindley distribution—recognized for its adaptability to diverse real-life datasets—the research develops statistical inference techniques to estimate distribution parameters with high precision and implements a two-sample prediction method to forecast unobserved failure times. Principal findings demonstrate that both maximum Likelihood and Bayesian estimators, supported by Markov Chain Monte Carlo methods, yield accurate parameter estimates, with Bayesian approaches showing slight superiority. Simulation results reveal reduced mean square errors and narrower credible intervals as sample sizes increase, while real-life applications to aircraft failure and leukemia survival data confirm the power Lindley distribution’s excellent fit. These results signify a robust framework for improving prediction accuracy under data constraints, offering significant advancements in reliability analysis and survival modeling. By providing a versatile methodology validated across industrial and clinical contexts, this study impacts statistical practice by equipping researchers with tools to address incomplete data challenges effectively, with broad implications for life-testing experiments in engineering, medicine, and beyond.

Keywords

Main Subjects

Files

Share

How to cite

Statistics