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: Nonlinear Regression STAT 563 Spring 2007 Regression Model • Recall that we can write the normal theory regression model as • Where x is a nvector of input variables, θ is a kvector of parameters, and the errors are independent N(0, σ 2 ) • The assumption that f(x , θ ) is a linear function of θ defines the linear regression model – For example, in equation (13.1), we can set θ = (β , β 1 ,..., β κ 29 • Linear regression is popular – Simplicity – Broad applicability – Elegant distributional and inferential results • What about when the expected value of the response is not a linear function of the parameters? – Results in nonlinear regression models ε θ + = ) , ( x f y Example • In our discussion on Forbes data, we looked at a theoretical relationship between the pressure and the boiling point of water as – Expected value function is not linear in parameters – We took logarithm to make it linear and verified that the assumptions of additive normal errors was reasonable in the transformed model + = BP PR 7 . 459 exp 1 γ γ Example • CobbDouglass function – Often used to develop economic models – Takes the form – Taking log on both sides will linearize the form – To fit the data (y,x) to a model with mean function specified above (*), we can use the linear model in (**) with α =ln( θ ) 1 1 ) , , ( θ θ θ θ x x f = ) ln( ) ln( ) , , ( ln( 1 1 x x f θ θ θ θ + = * ** ε θ α + + = ) ln( ) ln( 1 x y *** Example •...
View
Full
Document
This note was uploaded on 03/08/2009 for the course 960 563 taught by Professor Unknown during the Spring '07 term at Rutgers.
 Spring '07
 Unknown

Click to edit the document details