學(xué)術(shù)沙龍主題: Monitoring parameter change-points in lagged dependent regression models
報(bào)告人: 陳占?jí)郏嗪煼洞髮W(xué)教授
報(bào)告時(shí)間: 2025年4月19日(周六)；下午2:00—3:30
報(bào)告地點(diǎn)：南校區(qū)網(wǎng)絡(luò)安全大樓 121 會(huì)議室
邀請(qǐng)人：李本崇
報(bào)告人簡介：陳占?jí)郏嗪煼洞髮W(xué)研究生院院長，教授，博導(dǎo)，青海省統(tǒng)計(jì)學(xué)會(huì)副會(huì)長，入選青海省昆侖英才教學(xué)名師，青海省高校拔尖學(xué)科帶頭人等多個(gè)省級(jí)人才稱號(hào)，曾訪問加拿大英屬各類比亞大學(xué)和中科院系統(tǒng)所各1年；主要從事時(shí)間序列變點(diǎn)分析，小域估計(jì)、模型平均等方面的研究工作，先后主持國家自然科學(xué)基金3項(xiàng)，青海省自然科學(xué)基金6項(xiàng)；發(fā)表科研論文60余篇，出版學(xué)術(shù)專著1部，獲青海省自然科學(xué)獎(jiǎng)三等獎(jiǎng)1項(xiàng)。
報(bào)告摘要: This study addresses the change-point monitoring problem in time-dependent linear regression models, which incorporate autoregressive terms in both the response variable and the residuals. We propose a vector-valued CUSUM statistic based on the adaptive LASSO (ALASSO) algorithm to detect changes for the regression coefficients in the model. For the proposed test statistic, we establish the asymptotic distribution under the null hypothesis and demonstrate its divergence under alternatives. Through simulations, we show that our method significantly outperforms the classical CUSUM-based methods in terms of detection power and average run length, particularly when dealing with autocorrelated structures. Finally, we apply our monitoring procedure to a real data set to illustrate its effectiveness.
