6月27日 Don Rubin:Conditional Calibration and the Sage Statistician

时间:2016-06-26浏览:474设置

报告时间:6月27日上午10:00

报告地点:统计楼105报告厅

报告题目:Conditional Calibration and the Sage Statistician

报告人:Don Rubin(哈佛大学前系主任)

报告简介:

    Wise statisticians, when considering procedures to use, must be concerned with two general properties of those procedures.First, being approximately calibrated in the frequentist, Neymanian, sense of having appropriate long run operating characteristics; for example, 95% interval estimates should be confidence intervals in the sense of including the true values of their estimands at least 95% of the time (approximately) in repeated sampling no matter what the true value of the estimand.Second, sage statisticians should believe that in all problems they address, the intervals they provide are relatively accurate for the problems at hand; for example, their 95% intervals t should be interpretable as approximate Bayesian posterior intervals for their estimands under reasonable models.In general, simultaneously achieving both objectives is impossible.However, achieving conditional calibration is a realistic goal for sage statisticians.This perspective will be developed and discussed in this session.

返回原图
/