A test for heteroscedasticity applicable when the observations can be ordered according to non-decreasing variance. The test is carried out by omitting r central observations and running two separate ordinary least squares regressions on the first and the last (N − r)/2 observations. Under the null hypothesis of homoscedasticity the test statistic, GQ = S2/S1, has an F-distribution, with degrees of freedom p=q=(N−r)/2−K. Here S1 and S2 are the sums of squared residuals from the first and the second test regressions, N is the sample size, and K is the number of explanatory variables in the main regression. The optimum value of r is not known; r ≈ N/3 is frequently used.