This preview shows pages 1–13. Sign up to view the full content.
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 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 DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: C22.0103: Statistics for Business Control: Regression and Forecasting Hong Luo Section 003, Spring 2009 Tue/Thu/Fri, 11:00  12:15pm, Tisch 200 Stern School of Business New York University (Multiple) LINEAR REGRESSION Multiple Linear Regressions Model Building Multicollinearity Multiple Linear Regressions Model Building Multicollinearity Introduction to Multiple Linear Regression I A motivating example I The multiple linear regression model I Least squares estimates I Interpretation of the coefficients I Fitted value and residuals I R 2 and Adjusted R 2 A Motivating Example: gasoline consumption Shouldnt demand curve slopes downward? A Motivating Example: gasoline consumption I Theory suggests that demand for gasoline depends on price of gasoline and income level: G = f ( price , income ) I The upward slope seems to be the trace of equilibrium points along the supply curve. Q P D ( income 1 ) D ( income 2 ) D ( income 3 ) S I Micro 101 suggests that holding income constant , demand should be negatively correlated with price. Gasoline Consumption: US market, 1953  2004 Data File = newGas.MTP. Variables include: I GasQ: index number for gasoline physical consumption I GasP: index number for gasoline price I PCIncome: per capita real income I Pop: population in millions I PCGasQ: GasQ/Pop Gasoline Consumption: US market, 1953  2004 I All three variables are increasing over time. I Income is growing over time, which suggests that the main driver of gasoline consumption might be income not price The Simple Linear Regression: without controlling for income GasQPC = + 1 GasP + The coefficient of GasP is : when Price increases by 1 unit, consumption is expected to increase by 0.0016 units. Multiple Regression: holding income constant GasQPC = + 1 GasP + 2 PCIncome + I It looks like the theory works: controlling for income, we get negative price coefficient! The Multiple Linear Regression Model I Theory: y = f ( x 1 , x 2 , , x K ) I x 1 , x 2 , , x K are variables that may be useful in predicting y . I Theory specifies the model: there is a reason why certain variables appear in the equation. I The multiple linear regression model: y = + 1 x 1 + 2 x 2 + + K x K + I : the constant I k : the regression coefficient of x k . It represents the slope of the line relating y to x k when all the other x s are held fixed . I : the random noise, the collection of all factors that are not captured by x s. We make the same assumptions on as we did in the simple linear regressions. I is a RV with mean zero and variance 2 I is independent of x s The Multiple Linear Regression Model y = + 1 x 1 + 2 x 2 + + K x K + Again, the linear regression model is composed of two parts: I The systematic part: + 1 x 1 + 2 x 2 + + K x K I It accounts for the part of y that depends systematically on the explanatory variables ....
View
Full
Document
This note was uploaded on 11/07/2010 for the course ECON 0001 taught by Professor Kitsikopoulos during the Spring '08 term at NYU.
 Spring '08
 KITSIKOPOULOS

Click to edit the document details