活動起日：2014-12-29 發佈日期：2014-12-29 瀏覽數：894 2017-02-12 更新
時間: 1/9(五) 10:30-12:00
地點: 管理學院一號館2F 冠德講堂
演講人: Prof. Pinyuen Chen
講題: Heteroscedasticity: From Hypothesis Testing to Ranking & Selection
I first review two Stein-type two-sample procedures for selecting the largest normal mean under heteroscedasticity: (1) Bechhofer, Dunnett, and Sobel (1954, Biometrika) for common unknown variance and (2) Dudewicz and Dalal (1975, Sankhya) for uncommon unknown variances. Stein’s idea was to take an initial sample from each population for variance estimation and then to utilize the estimated variance from the first sample in designing a second sample for testing.
Then I recall two procedures for comparing populations with a standard/control for the case of known variances: (1) Bechhofer & Turnbull (1976, JASA)’s approach in comparing normal means with a standard and (2) Thall, Simon, & Ellenberg (1988, Biometrika)’s two-stage selecting-and-testing approach in selecting the best treatment among several experimental treatments, provided that it is better than a control. In extending the results of above two procedures to heteroscedastic case, we apply Stein-type two-sample concept. Simulation comparisons will be presented to illustrate the proposed methodologies.
This talk is at an introductory level that is accessible to every student who has taken one mathematical statistics course.