This preview shows pages 1–7. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Statistics 191: Introduction to Applied Statistics Model Selection Jonathan Taylor Department of Statistics Stanford University March 1, 2010 1 / 1 Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Topics Outline Goals of model selection. Criteria to compare models. (Some) model selection. Bias variance tradeoff. 2 / 1 Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Election data Description Variable Description V votes for a presidential candidate I are they incumbent? D Democrat or Republican incumbent? W wartime election? G GDP growth rate in election year P (absolute) GDP deflator growth rate N number of quarters in which GDP growth rate > 3 . 2% 3 / 1 Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Model selection Problem & Goals When we have many predictors (with many possible interactions), it can be difficult to find a good model. Which main effects do we include? Which interactions do we include? Model selection procedures try to simplify / automate this task. Election data has 2 6 = 64 different models with just main effects! 4 / 1 Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Model selection General comments This is an unsolved problem in statistics: there are no magic procedures to get you the best model. In some sense, model selection is data mining. Data miners / machine learners often work with very many predictors. Our model selection problem is generally at a much smaller scale than data mining problems. 5 / 1 Statistics 191: Introduction to Applied Statistics Jonathan Taylor Department of Statistics Stanford University Model selection Hypothetical example Suppose we fit a a model F : Y n 1 = X n ( p +1) ( p +1) 1 + n 1 with predictors X 1 ,..., X p ....
View
Full
Document
This document was uploaded on 03/16/2010.
 Winter '09
 Statistics

Click to edit the document details