This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Section 6 - Econ 140 GSIs: Hedvig, Tarso, Xiaoyu * 1 Review: Multiple Regression (continued) 1.1 Set-up (reminder) Y i = β + β 1 X 1 i + β 2 X 2 i + .. + β K X Ki + u i , (1) where β k is the partial (marginal) e ect of X k on Y , in other words, it shows the change in Y as a result of a one unit increase in X k holding all other regressors constant ( ceteris paribus e ect). You can think about the intercept as being the slope coe cient of a regressor X i , which takes the value of 1 for all observations. 1.2 Goodness-of- t in the MLR model R 2 : fraction of the sample variance of Y i explained (or predicted by) the regressors. Formula: R 2 = 1- SSR TSS In multiple regression, the R 2 increases whenever a regressor is added, unless the estimated coef- cient on the added regressor is exactly zero. Adjusted R 2 (or ¯ R 2 ): modi ed version of the R 2 that does not necessarily increase when a new regressor is added. Formula: ¯ R 2 = 1- N- 1 N- K- 1 SSR TSS = 1- s 2 ˆ u s 2 Y Three useful things about ¯ R 2 : 1. N- 1 N- K- 1 is always greater than 1 ⇒ ¯ R 2 always less than R 2 2. Adding a regressor has two opposite e ects on ¯ R 2 : SSR falls, which increases it, but N- 1 N- K- 1 increases, which decreases it. 3. ¯ R 2 can be negative: this happens when the regressors, taken together, reduce the sum of squared residuals by such a small amount that this reduction fails to o set the factor N- 1 N- K- 1 . SER = q SSR N- K- 1 , where K is the number of explanatory variables and -1 stands for the constant. * Many thanks to previous GSIs, Edson Severnini and Raymundo M. Campos-Vazquez, as this note is based on theirs. All errors are ours. 1 1.3 Multicollinearity 1.3.1 De nition Presence of linear correlation between explanatory variables. 1.3.2 Imperfect multicollinearity Imperfect multicollinearity arises when one of the regressors is very highly correlated - but not perfectly correlated - with the others regressors. In other words, there is a linear function of the regressors that is highly correlated with another regressor. Consequence : higher SE ( ˆ β OLS ) ⇒ it is easier to fail to reject the null hypothesis. Remark : The imperfect multicollinearity does not pose any problems for the theory of the OLS esti- mators. Indeed, a purpose of OLS is to sort out the independent in uences of the various regressors when these regressors are potentially correlated. 1.3.3 Extreme case: perfect collinearity Perfect collinearity arises when one of the regressors is perfectly correlated with the others regressors. In other words, one regressor can be expressed as a linear function of the others. Consequence : Let Y i = β + β 1 X 1 i + β 2 X 2 i + ... + β k X ki + u i be the regression model. Now, suppose for example that X 1 i = 1 + X 2 i + X 3 i + ... + X ki : ⇒ Y i = β + β 1 (1 + X 2 i + X 3 i + ... + X ki ) + β 2 X 2 i + ... + β k X ki + u i ⇒ Y i = ( β + β 1 ) + ( β 1 + β 2 ) X 2 i + ... + ( β 1 + β k ) X ki + u i ⇒ Y i = γ...
View Full Document
This note was uploaded on 02/02/2012 for the course ECON 140 taught by Professor Duncan during the Spring '08 term at Berkeley.
- Spring '08