LOGIN ENGLISH

제목 [학술세미나] [학과세미나] 9월 25일(금) 학과세미나 안내
작성일 2020-09-18 08:41:22
내용 [학과세미나] 9월 25일(금) 학과세미나 안내

▪제목 : Multiple response logistic regression with applications to cancer cell-line encyclopedia data
▪연사 : 박세영 (성균관대 통계학과)
▪일시 : 2020년 9월 25일(금) AM 11:00 – 12:00
▪진행 : ZOOM 활용 비대면 세미나
        (접속 링크 : https://snu-ac-kr.zoom.us/j/94500517398)


초  록
  We consider multiple response logistic regression models in which the number of predictors increases with the sample size. Motivated by the analysis of cancer cell-line encyclopedia data, we assume that the underlying regression coefficient matrix is low rank. We introduce a penalized logistic regression using a weighted nuclear norm regularizer. For statistical properties, we derive the non-asymptotic error bounds and prove the structural consistency of the proposed estimator, which guarantees a rank consistency of the estimator. We also propose a new information criterion to choose a regularization parameter of the proposed method and show that it selects the true rank of the underlying coefficient matrix. We demonstrate the improvement in estimation accuracy by considering multiple responses jointly through the proposed estimator when the underlying coefficient matrix is indeed low rank. In our analysis of both simulated data and the cancer cell line encyclopedia data, the proposed method can better predict drug responses than the existing methods.
파일 세미나 안내_200925_박세영.hwp [14.5KB]