G26/27/29B Econometrics, with Trade/Development App more

TIANX_Handout2 for ECON G27/9B:09/10 Diagnostic Tests in PcGive ♦ AR test: test for Error Autocorrelation H0: cov ( ut us ) = 0, t ≠ s ♦ ARCH test: test for AutoRegressive Conditional Heteroscedasticity (see Engle, 1982) H0: E ut2 ut −1, ,ut − r  = σ 2   ♦ Normality test: test for Normality H0: ut ~ N ( 0, σ 2 ) ♦ Hetero test: Heteroscedasticity test using squares (not for ML, see White, 1980) H0: Var ( ut ) = E [ut ] = σ 2 - unconditional homoscedasticity ♦ Hetero-X test: Heteroscedasticity test using squares and cross-products (not for ML, see White, 1980) H0: errors are homoscedastic Note, it is only calculated if there is a large number of observations relative to the number of variables in the regression, i.e. T k ( k + 1) ♦ RESET test: test for Regression Specification/functional form (OLS only, see Ramsey, 1969) ˆ H0: No mis-specification of the model. i.e. No powers of dependent variable yt ˆ ˆt such as ( y t2 , y 3 …) have been omitted. Example test summary from Exercise 1, data: Exrate Question: Can you identify the problems of the model? How would you make improvements?