Different tests

Question 1

How do we test for fixed and random effect models?

Answer 1

Use the Hausman test: If estimates are very different then the RE is inconsistent and the FE is more appropriate

Question 2

How do we test if we need to use an IV?

Answer 2

Use the Durbin-Wu-Hausman test

Question 3

What are the 3 tests for heteroskedasticity?

Answer 3

GQ test: When we are confident on the source… SSR3/SSR1, n3-k-1, n1-k-1, null hypothesis is that there is no heteroskedasticity
Breusch-Pagan test: Not confident on the source and a small test… run the equation, obtain residuals and square them, test statistic nR^2
White test: larger sample, do the same but regress squared residuals vs explanatory variables and square their cross products, same test as before

Question 4

How do we test for autocorrelation normally?

Answer 4

Use a Durbin-Watson test: upper and lower bounds

Question 5

How do we test for individual or joint sources of autocorrelation?

Answer 5

T/F-test

Question 6

How do we test for autocorrelation for an equation with a lagged dependent variable?

Answer 6

We use Durbin’s h test: We need the variance of the coefficient of the lagged variable, number of observations and estimate of p (rho) = 1-0.5d