Lec11.ModelSelection - Model Selection KNNL Chapter 9...

Info iconThis preview shows pages 1–11. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Model Selection KNNL Chapter 9 (9.1-9.3) Building the Regression Model I Choosing the Best Model Best Fitting (Smallest Sum of Squares) Most Parsimonious (Fewest Parameters) Best Predictor Most Helpful (Predictive Sense) Easiest to Interpret Easiest to Modify and Update Cheapest Nicest Color alligator$Length 60 80 100 120 140 100 200 300 400 500 600 alligator$Weight Best Model Analysis of Variance 1- n ) ( Total MSE MSR SSE MSE p- n ) ( Error 1 MSR 1- p ) ( Regression Test Square Mean df squares of Sum Source 2 2 2 - = =- =- =- =- = Y Y SSTO Fstat p n Y Y SSE p SSR Y Y SSR i i i i F-test MSE MSR df SSE df SSR df SSE df df SSE SST F df RSS df df RSS RSS F E R E E = =-- =-- = T Proposed Proposed Proposed Prior Proposed Prior General Null Model vs. Full Model Information Content of Data Watch for Highly Correlated Data Watch for Contrast in Data Tavern Customers Keys i i i i i i i X Y X X if X X Y 1 2 1 2 1 2 2 1 1 ) ( + + = + + = Importance of Data A well designed experiment is extremely useful A well designed survey can be as useful Observational studies may be all that is available or affordable and may be quite informative, but let the user beware Keep your design hat on Surgical Unit Data A hospital surgical unit was interested in predicting survival in patients undergoing a particular type of liver operation Random selection of 108 patients Predictor variables gathered from a pre- operation evaluation Response: Post-operation survival time in days Surgical Unit Data Surgical = read.table("CH09TA01.txt") # names(Surgical) = c( "BloodClot.Score", "Prognostic.Index", "Enzyme.Test", "Liver.Test", "Age", "Gender", "AlcUse.Mod", "AlcUse.Hvy", "Survival.Time", "LogSurv.Time") Surgical Surgical[1:5,] BloodClot.Score Prognostic.Index Enzyme.Test Liver.Test Age 1 6.7 62 81 2.59 50 2 5.1 59 66 1.70 39 3 7.4 57 83 2.16 55 4 6.5 73 41 2.01 48 5 7.8 65 115 4.30 45 Gender AlcUse.Mod AlcUse.Hvy Survival.Time LogSurv.Time 1 0 1 0 695 6.544 2 0 0 0 403 5.999 3 0 0 0 710 6.565 4 0 0 0 349 5.8544 0 0 0 349 5....
View Full Document

Page1 / 43

Lec11.ModelSelection - Model Selection KNNL Chapter 9...

This preview shows document pages 1 - 11. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online