Publication: ENHANCING INFERENCE FOR RAMA DISTRIBUTION: CONFIDENCE INTERVALS AND THEIR APPLICATIONS
1
0
Issued Date
2025-01-01
Resource Type
ISSN
0256422X
Scopus ID
2-s2.0-105017450049
Journal Title
Journal of Statistical Research
Volume
59
Issue
1
Start Page
107
End Page
129
Rights Holder(s)
SCOPUS
Bibliographic Citation
Journal of Statistical Research Vol.59 No.1 (2025) , 107-129
Suggested Citation
Panichkitkosolkul W., Khruachalee K., Volodin A. ENHANCING INFERENCE FOR RAMA DISTRIBUTION: CONFIDENCE INTERVALS AND THEIR APPLICATIONS. Journal of Statistical Research Vol.59 No.1 (2025) , 107-129. 129. doi:10.3329/jsr.v59i1.83689 Retrieved from: https://hdl.handle.net/20.500.14740/50581
Author(s)
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
This research introduces and investigates four approaches for constructing confidence intervals (CIs) associated with the parameter of the Rama distribution—a model often applied in lifetime data modeling. The methods under consideration comprise the likelihood-based, Wald-type, bootstrap-t, and bias-corrected and accelerated (BCa) bootstrap intervals. To assess their practical utility, both Monte Carlo simulations and real data applications were utilized, emphasizing key performance indicators such as empirical coverage probability (ECP) and average width (AW) under various experimental conditions. To improve computational efficiency, a closed-form expression for the Wald-type CI was formulated. Simulation findings indicated that, across most situations, the ECPs obtained from both the likelihood-based and Wald-type CIs remained closely aligned with the nominal 95% confidence level. However, when the sample size was small, both the bootstrap-t and BCa bootstrap CIs yielded ECPs that fell short of the nominal level. As the sample size increased, the ECPs associated with these methods progressively approached the targeted confidence level, though variations in parameter values continued to influence their performance. The practical utility of these CIs was further validated through their application to two real-world datasets: monthly tax revenue in Egypt and plasma concentrations of indomethacin. The results from these applications were consistent with the findings of the simulation study, confirming the robustness and applicability of the proposed methods.