講座編號：jz-yjsb-2022-y017
講座題目：Activation discovery with FDR control Application to fMRI data
主 講 人：王兆軍 教授 南開大學
講座時間：2022年5月11日（星期三）下午14:00
講座地點：騰訊會議，會議ID:518 222 217
參加對象：數學與統計學院全體教師及研究生
主辦單位：數學與統計學院、研究生院
主講人簡介：
王兆軍，南開大學統計與數據科學學院教授、博士生導師、執行院長和黨總支書記，統計研究院院長，中國現場統計研究會副理事長，中國工業統計教學研究會副會長，天津數據科學與技術學會理事長，天津市學位委員會數學與統計學科評議組召集人，曾獲國務院政府特殊津貼，全國百篇優博指導教師，教育部全國高校自然科學獎二等獎及天津市自然科學獎一等獎。王兆軍教授的主要研究方向包括統計過程控制(SPC)、非(半)參數回歸、降維、高維數據分析、變點等，已在Journal of the American Statistical Association、Annals of Statistics、Biometrika、Statistica Sinica等專業頂級期刊上發表高質量學術論文110余篇，先后主持國家自然科學基金重點項目、面上項目、教育部博士點基金項目等10余項，現擔任Statistical Theory and Related Fields、《統計信息論壇》、《數學進展》等雜志編委和《數理統計與管理》等雜志副主編。
主講內容：
Data arriving in “streams” from a large number of sources is ubiquitous, a portion of which usually incurs structural changes during the time-course of data acquisition. For example, in fMRI analysis, some brain regions become active associated with task-related stimuli or even in resting-states. Such a region corresponds to an activated data stream. We are aiming to measure the uncertainty of discovering data streams in activation via the tool of the false discovery rate (FDR). Borrowing ideas from recent developments of the FDR control methodologies, we propose a simple yet effective method to achieve this purpose meanwhile taking unknown asynchronous change patterns and spatial dependence into consideration. Its validity on controlling the FDR is justified by asymptotic analysis. Numerical experiments indicate that the proposed method is both accurate and powerful. It is also applied in a real fMRI data analysis. A R package SLIP is developed to implement the proposed method.