Searching for important variables in large data bases: Multiple testing and model selection

Malgorzata Bogdan

  • Date:

    14 MAY
    -
    17 MAY 2019
     at 10:00

  • Event location: Seminar Room, 1st Floor, Department of Statistis Science, Via Belle Arti 41, Bologna

  • Type: Cycle 34 - Short courses and seminars

14 May     10-13
15 May     14-17
16 May     10-13
17 May     10-13

Abstract

During this course we will discuss the problem of identifying important predictors in large data bases. We will start from the generic problem of multiple testing and then cover the problem of identifying important predictors through model selection criteria and regularization methods (like LASSO or SLOPE). We will also present knockoff methodology for constructing control variables, which allow to estimate the level of noise in the estimates of regression coefficients. The lectures will be supplemented with computer labs in R, where student will have a chance to verify the statistical properties of discussed methodology using computer simulations.