سمینار تخصصی گروه آمار

سخنران: خانم دكتر سولماز سيف اللهي

زمان: چهارشنبه 16ام مهر 1404 ساعت 10 صبح

مکان : سالن دكتر بزرگ نيا

Bayesian approach in restricted regression Models

Abstract:

In statistical modeling, incorporating prior knowledge about model parameters can greatly improve both the efficiency and interpretability of inference. However, most Bayesian approaches assume that prior information can be expressed through simple linear equalities. In reality, parameters often follow inequality relationships that reflect underlying physical, economic, or logical constraints. Accounting for these relationships leads to models that better represent real-world systems and behave more consistently with scientific expectations.

In this work, we introduce a Bayesian framework for analyzing regression models with linear inequality constraints on parameters. We develop efficient computational algorithms for performing posterior inference under these restrictions and show how Bayesian variable selection can be integrated to build more parsimonious and interpretable models. This approach offers a flexible and principled way to handle constrained inference, providing a solid foundation for analyzing complex models where prior structural knowledge matters.