# MIAMI IES 612 - Lecture Notes (22 pages)

Previewing pages 1, 2, 21, 22 of 22 page document
View Full Document

## Lecture Notes

Previewing pages 1, 2, 21, 22 of actual document.

View Full Document
View Full Document

## Lecture Notes

92 views

Lecture Notes

Pages:
22
School:
Miami University
Course:
Ies 612 - Environmental Analysis&Modelng
##### Environmental Analysis&Modelng Documents
• 15 pages

• 6 pages

• 11 pages

• 26 pages

• 11 pages

• 11 pages

• 2 pages

• 21 pages

• 21 pages

• 11 pages

• 29 pages

• 6 pages

• 18 pages

• 7 pages

• 7 pages

• 3 pages

• 12 pages

• 16 pages

• 6 pages

Unformatted text preview:

00 09 Monday January 14 2019 1 IES 612 STA 4 573 STA 4 576 Spring 2005 Week 05 IES612 lecture week05 doc Model Building Comments Brain first computer second pick variables for inclusion in the model that are important based on your knowledge of the system This doesn t exclude the possibility of exploratory analyses it simply asks for careful consideration of variables to be considered for inclusion in a model Hierarchical Principle If you have a higher order term include all lower order constituent terms For example X2 should be in a model if X the linear term is already present The interaction term X1 X2 should be in the model if X1 and X2 are in the model Aside Interactions can be used to allow for the relationship between the response variable Y and an X variable say X1 to vary with levels of another predictor variable X2 We will discuss this in more detail when we consider factorial designs in anova models SYNERGY observed response is beyond what would be predicted from the additive effect of the two variables e g lung cancer as a function of smoking and asbestos exposure ANTAGONISM observed response is less than would be predicted from the additive effects of the two variables e g competition for the same binding sites Principle of Parsimony The model should be no more complicated than required to describe the response Variable selection methods i ALL POSSIBLE REGRESSIONS SAS Proc RSQUARE Suppose you have 5 possible predictor variables X1 X2 X3 X4 X5 then you have 25 1 31 possible models variables present 1 X1 X2 X1 X2 X3 X4 X5 X3 00 09 Monday January 14 2019 2 X4 X5 2 X1 X1 X1 X1 X2 X3 X4 X5 X2 X2 X2 X3 X4 X5 X3 X3 X4 X4 3 X1 X1 X1 X1 X1 X1 X2 X2 X2 4 5 X1 X1 X1 X1 X1 X3 X4 X5 X3 X3 X2 X2 X2 X2 X2 X2 X2 X2 X5 X5 X3 X3 X3 X3 X3 X3 X3 X3 X4 X4 X4 X4 X4 X4 X4 X4 X4 X4 X5 X5 X5 X5 X5 X5 X5 X5 X5 X5 Which model do you select 1 Models with smallest incremental increase in R2 or smallest incremental decrease in SSE look for the elbow in these indices 2 Adjusted R2 1

View Full Document

Unlocking...