A0251
Title: Exact statistical inference for some transformed gamma distributions: A generalized inference approach
Authors: Bowen Liu - University of Missouri - Kansas City (United States) [presenting]
Abstract: The transformed gamma distributions are frequently used for modeling extreme events across diverse fields, including hydrology, meteorology, and the insurance industry. Although these distributions demonstrate strong goodness-of-fit characteristics, exact inference for extreme quantiles remains challenging. Currently available parametric methods for quantile inference typically rely on approximation or bootstrapping techniques, which often demonstrate poor performance, particularly when sample sizes are relatively small. An exact inferential approach is presented for several transformed gamma distributions using generalized inference methodology. Additionally, the relationship is explored between the generalized inference method developed by Weerahandi and generalized fiducial inference. To evaluate the performance of the exact inference procedure, a series of simulation studies are conducted. Results from these simulations indicate that the exact inference method consistently outperforms approximation-based and bootstrapping methods across multiple scenarios. Moreover, the practical applicability of the method was validated using multiple real-world datasets.
