Your browser does not support JavaScript!
2017-03-10(五),主講人:Prof. Shinpei Imori (Graduate School of Engineering Science, Osaka University)


統 計 學 研 究 所

專 題 演 講


講 題: Auxiliary variable selection by information criteria
演講者: Prof. Shinpei Imori (Graduate School of Engineering Science, Osaka University)
時 間: 106年03月10日(星期五)10:40 - 12:00noon  (10:20 - 10:40am 茶會於統計所821室舉行)
地 點: 綜合三館837室
摘 要:


An aim of statistical modeling is to obtain a model, which fits testing data (future observations). In a basic statistical analysis, variables between the training and testing data are assumed to be of the same kind when the model is constructed. Hence, if some variables in the training data are not observed in the testing data, the constructed model cannot be directly applied to the testing data. On the other hand, it is possible that some variables are observed only in the training data due to some reason. In the present paper, such variables are referred to as the auxiliary variables, and we use the auxiliary variables to construct a statistical model. Utilizing the auxiliary variables sometimes improves the prediction accuracy in the testing data but sometimes worsens, depending on the situation. We propose information criteria in order to select the useful auxiliary variables.


敬請公佈                                 歡迎參加

  • 友善列印
  • 分享
Voice Play