Econometrics-I-14

# Econometrics-I-14 - Applied Econometrics William Greene...

This preview shows pages 1–7. Sign up to view the full content.

Applied Econometrics William Greene Department of Economics Stern School of Business

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

View Full Document
Applied Econometrics 14. Nonlinear Regression and Nonlinear Least Squares
Nonlinear Regression What makes a regression model “nonlinear?” Nonlinear functional form?     Regression model: y i   =  f(  x i  ,  β  )  +   ε i Not necessarily:          y i  = exp( α ) +  β 2 *x i  +  ε i                                 β  = exp( α )                                y i   = exp( α )x i β exp( ε i )     is “loglinear” Models can be nonlinear in the functional form of the  relationship between y and x, and not be nonlinear for  purposes here. We will redefine “nonlinear” shortly, as we proceed.

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

View Full Document
Nonlinear Least Squares Least squares: Minimize wrt  β  ½  Σ i  [y i   - f( x i , β )] 2                                           = ½  Σ i  e i 2 First order conditions:     ½ Σ i [y i - f( x i , β )] 2  ]/ ∂β             = ½ Σ i (-2)[y i - f( x i , β )]  f( x i , β )/ ∂β           =  - Σ i  e i   x i 0   0 (familiar?) There is no explicit solution,  b  = f( data ) like LS. (Nonlinearity of the FOC defines nonlinear model)
Example How to solve this kind of set of equations:  Example,            y i   =   β 0 +  β 1x i β 2  +  ε i .            [ ½ Σ i  e i 2 ]/ ∂β 0  =  Σ i  (-1) (y i  -  β 0 -  β 1x i β 2 )  1             =  0            [ ½ Σ i  e i 2 ]/ ∂β 1  =  Σ i  (-1) (y i  -  β 0 -  β 1 x i β 2 ) x i β 2           =  0            [ ½ Σ i  e i 2 ]/ ∂β 2  =  Σ i  (-1) (y i  -  β 0 -  β 1 x i β 2 β 1 x i β 2 lnx i  =  0 Nonlinear equations require a nonlinear solution.  We’ll return to that  problem shortly. This defines a nonlinear regression model.  I.e., when the first order  conditions are not linear in  β . (!!!) Check your understanding.  What does this produce if f(  x i  ,  β  ) =  x i ′ β ?  (I.e., a linear model)

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

View Full Document
The Linearized Regression Model Linear Taylor series:    y  = f( x i , β )  +   ε .    Expand the regression around some point,  β 0 . f(
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 11/23/2011 for the course ECON B30.3351 taught by Professor Professorw.greene during the Spring '10 term at NYU.

### Page1 / 24

Econometrics-I-14 - Applied Econometrics William Greene...

This preview shows document pages 1 - 7. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online