ETC2410 Lecture Notes - Lecture 8: Heteroscedasticity, Conditional Expectation, Scatter Plot
Document Summary
Definition of heteroskedasticity and its consequences for ols. Testing for heteroskedasticity: breusch-pagan test, white test. Weighted least squares when heteroskedasticity is known up to a multiplicative constant. If we detect htsk, we can conduct reliable inference based on the ols estimator, or even obtain a more efficient estimator: detecting htsk is possible by looking at a scatter plot, but only if there is one x. If we had ui, then we could square it and estimate the conditional expectation function of ui. Ols is still unbiased, it is no longer blue and ols standard errors are unreliable: practical problem is that t- and f-statistics based on ols standard errors are unusable. Solution 2: transform the model: logarithmic transformation of y. If population model has log(y) as the dependent variable but we have used y, this misspecification can show up as htsk errors.