Bayesian Point and Interval Prediction of Ordered Observations in Future Censored Samples from Contagious Geometric Distribution

SYNOPTIC ABSTRACT

This article presents the study of prediction analysis on future ordered sample observations from a two-component contagious model of geometric distribution using a Bayesian approach. To be more specific, a theoretical methodology is proposed to get approximate point and interval prediction of ordered statistics in the future samples when these samples are censored by right type-I, right type-II, and left censoring techniques. Point predictors are developed using a symmetric and an asymmetric loss function. To show the usefulness of the derived mathematical results, the study applies them to real-world data and considers, for illustration, the computation of point prediction for quartiles and 95% prediction intervals for median and inter-quartile range in the future censored samples.

KEYWORDS AND PHRASES:

• Bayesian prediction
• geometric distribution
• mixture model
• loss functions
• ordered observation
• censoring schemes