Submitted by: Submitted by tiandijie333
Views: 202
Words: 3517
Pages: 15
Category: Business and Industry
Date Submitted: 05/02/2012 11:55 PM
TOPIC 02 MULTIPLE REGRESSION ANALYSIS
Wang Jia
FIN401 Financial Modeling Macau University of Science and Technology September 2009
OUTLINE
1. Motivation for Multiple Regression 2. The Linear Regression Model (with Many Regressors) 3. Recap 4. Statistical Properties of OLS
Wang Jia
Topic 2: Multiple Regression Analysis
September 2009
2
1. Motivation for Multiple Regression
MOTIVATION FOR MULTIPLE REGRESSION
Will our innovations increase is we:
grow more? employ more people? spend more money on R&D?
You consider models:
log ( y ) = β 0 + β1lsales + ε log ( y ) = α 0 + α1employ + υ log ( y ) = γ 0 + γ 1lR & D + θ
Problem: error correlated with explanatory variables in all these models
Wang Jia Topic 2: Multiple Regression Analysis September 2009 3
OUTLINE
1. Motivation for Multiple Regression 2. The Linear Regression Model (with Many Regressors)
1. Obtaining the OLS estimates 2. Interpretation of multiple regression 3. Goodness of fit
3. Recap 4. Statistical Properties of OLS
Wang Jia
Topic 2: Multiple Regression Analysis
September 2009
4
2. The Linear Regression Model (with many regressors)
THE LINEAR REGRESSION MODEL
There is an (unknown) relation in the population:
y = f ( x1 , x2 , … , xk ) + u
Steps:
Collect sample Assumptions on sampling process and on y = f x1 , x2 , … , xk + u Obtain an estimate of y = f x1 , x2 , … , xk which has desirable properties under these assumptions Hopefully, the new assumptions are more likely to hold!!!
(
)
(
)
Wang Jia
Topic 2: Multiple Regression Analysis
September 2009
5
2. The Linear Regression Model (with many regressors)
THE LINEAR REGRESSION MODEL
As. 1: We assume that the population relation is linear:
y = β 0 + β1 x1 + β 2 x2 +… + β k xk + u
x j = independent variable j y = dependent variable u = error term β0 = intercept
Dy βj = slope of variable j Dx j
Wang Jia
Topic 2: Multiple Regression Analysis...