click below
click below
Normal Size Small Size show me how
Regression
Tests
Question | Answer |
---|---|
What is the basic idea of GLS? | 1) Find rho by regressing lagged residuals against original residuals 2) Correct for serial correlation by subtracting lagged variables from original variables (and subtract rho from one on intercept) to remove serial correlation out of the equation. |
What are the 3 requirements to use DW test? | 1) Must have an intercept 2) Cannot have lag regressors 3) Must be 1st order serial correlation |
How do you use the DW test? (3) | 1) Null is that serial correlation exists 2) DW range is 0-4; 0 = extreme serial correlation 3) benchmark is 2 (if DW is greater than 2, no serial correlation) • if DW is < dL, serial correlation exists • if DW > dU, no serial correlation exists |
What are the 2 steps in the Park Test? | 1) If you know Z factor, run regression of lnZ as independent variable and lne2 as dependent variable 2) Test significance of Z coefficient – if significant, you know that Z is the cause of the heteroskedasticity |
What are the 2 steps in the White Test? | 1) Runs regression of the square and product of all variables against square of residuals 2) Test overall significance with chi-squared test (N* ) - If chi test is significant, there is heteroskedasticity - White can detect heteroskedasticity, but can' |
What is a benefit of the White Test? | Can also test for complex heteroskedasticity |