Term
| What is homoscedasticity? |
|
Definition
| It refers to the variance of the error term, conditioned on the regressors, is constant: Var(ui|x1...xk)=σ2 |
|
|
Term
|
Definition
If the variances do not remain constant, we encounter heteroscedasticity
[image] |
|
|
Term
| What about our model is nullified by heteroscedasticity? |
|
Definition
- The OLS estimators are still unbiased and consistent
- However, the standard errors are biased which makes the t- and F-statistics no longer valid
- The coefficient estimates are no longer usable
- OLS model is no longer BLUE
|
|
|
Term
Robust Standard Errors
(Huber-White standard error) |
|
Definition
| sqrt{Avar(ßj)=[∑rij2ui2/RSSj2]} |
|
|
Term
| How does testing for heteroscedasticity work? |
|
Definition
| test whether the squared error terms are related to one or more of the regressors |
|
|
Term
| What is a famous test for heteroscedasticity? |
|
Definition
| Breush-Pagan test for heteroscedasticity |
|
|
Term
| First line of defense for heteroscedasticity |
|
Definition
|
|
Term
| If we know something more specific about the heteroscedasticity, we can find more efficient estimators than OLS |
|
Definition
Solve with Weighted Least Squares (WLS)
Var(u|x)=σ2 h(x)
where h(x) is a functional form of the regressors
|
|
|
