ارسل ملاحظاتك

ارسل ملاحظاتك لنا







Simulation Study for the Bayesian Reciprocal Adaptive Lasso Quantile Regression

المصدر: مجلة القادسية للعلوم الإدارية والاقتصادية
الناشر: جامعة القادسية - كلية الادارة والاقتصاد
المؤلف الرئيسي: Lafta, Zahraa Awaid (Author)
مؤلفين آخرين: Flaih, Ahmad Naeem (Co-Author)
المجلد/العدد: مج24, ع4
محكمة: نعم
الدولة: العراق
التاريخ الميلادي: 2022
الصفحات: 249 - 264
ISSN: 1816-9171
رقم MD: 1345597
نوع المحتوى: بحوث ومقالات
اللغة: الإنجليزية
قواعد المعلومات: EcoLink
مواضيع:
كلمات المؤلف المفتاحية:
Reciprocal Adaptive Lasso | Gibbs Sampler | Simulation | Full Conditional Posterior Distribution
رابط المحتوى:
صورة الغلاف QR قانون
حفظ في:
المستخلص: Regression analyses have two major purposes, explanation and the prediction. The explanation concept of the regression model can be capture by introducing the more interpretable model via variable selection procedure, while the prediction ability of the regression model can be capture by balancing between the bias and the variance of the interest parameter estimates. This paper explores the Bayesian adaptive lasso method, a shrinkage method that provides estimation and variable selection procedure, also this method yields more interpretable model with more prediction accuracy. The reciprocal lasso has favorable properties comparing with lasso and for that we utilize two scale mixture formulation, the first one is the scale mixture of normals and the scale mixture of uniforms. New hierarchical prior model and full conditional posterior distribution with the Gibbs sampler algorithm have developed. Two simulation scenarios conducted to test the performance of the proposed Bayesian methods in quantile regression. The results explained that the proposed models are comparable with some methods.

ISSN: 1816-9171

عناصر مشابهة