* Use VOTE1.dta
regress voteA prtystrA democA lexpendA lexpendB
* Obtain OLS residuals
predict res, residual
* This just shows that the residuals are, by the Least Squares procedure,
* always orthogonal (independent of) the explanatory variables, X'uhat = 0
regress res prtystrA democA lexpendA lexpendB
generate res2 = res^2
* The Breusch-Pagan Test for Heteroskedasticity (See equation (8.19), p. 253)
* Use the overall F-statistic from this equation 
regress res2 prtystrA democA lexpendA lexpendB
* The special case of White's test (See equation (8.20), p. 253)
regress voteA prtystrA democA lexpendA lexpendB
* Obtain predicted values of voteA = voteA_hat
predict voteA_hat
generate voteA_hat2 = voteA_hat^2
* Use the overall F-statistic from this equation
regress res2 voteA_hat voteA_hat2