kaiyun開(kāi)云官方網(wǎng)站“創(chuàng)源”大講堂研究生學(xué)術(shù)講座

講座題目：Variable selection via additive conditional independence 

報(bào)告人：    Li, Bing教授，賓州州立大學(xué)統(tǒng)計(jì)系

主持人：    殷向榮教授
講座時(shí)間：  2014年6月27日上午11:00

講座地點(diǎn)：  犀浦校區(qū)二號(hào)教學(xué)樓X2511

內(nèi)容簡(jiǎn)介： We propose a variable selection method for regression with high-dimensional predictors,  which does not rely on any   regression model or predictor distribution. More importantly, unlike many existing variable selection methods, which target the mean of the response alone, the proposed method targets a set of attributes of the response, such as its mean, its variance, or its entire distribution. The proposed variable selector is based on a new statistical relation, called additive conditional independence, that was introduced recently for graphical models (Li, Chun, and Zhao, 2013). We establish the estimation consistency and convergence rate, as well as variable-selection consistency of the proposed method. Through simulation comparisons we demonstrate that our method performs better than several existing methods when the predictor affects several attributes (such as mean and variance) of the response, and  performs competently in the classical setting where the predictors affects the mean alone. We apply the new method to a data set concerning how genetic traits affect the weights of female mice.

主講人簡(jiǎn)介：

      Li，Bing，賓州州立大學(xué)統(tǒng)計(jì)系教授，博士畢業(yè)于美國(guó)芝加哥大學(xué)，現(xiàn)任國(guó)際統(tǒng)計(jì)雜志Annuals of Statistics, Journal of Statistical Planning and Inference和Statistica Sinica副主編。在Ann. Statist.、JASA、Biometrika、Statist. Sinica、Canad. J. Statist.、Scandinavian Journal of Statistics, JSPI等國(guó)外頂尖期刊發(fā)表學(xué)術(shù)論文五十余篇。完成美國(guó)國(guó)家自然科學(xué)基金項(xiàng)目七項(xiàng)，現(xiàn)主持一項(xiàng)。

主辦：研究生院
承辦：kaiyun開(kāi)云官方網(wǎng)站