Abstract
Goldfeld-Quandt test is one of the three most popular tests to detect problem of heteroscedasticity in regression analysis. This test was proposed by Goldfeld and Quandt in 1965. It is implemented as follow: (1) record the data set according to the values of the independent variable, which is suspected to be cause of heteroscedasticity, from lowest to highest, (2) divide the sample size, n, that was already sorted to three parts and omit the middle part with size c. Thus, we obtain two subsamples of sizes n1 and n2 , usually n1 n2 (n c) 2 . If heteroscedasticity is present, then the variance of the last subsample will not be the same as the variance of the first subsample; it tends to be larger. The F-test for the ratio of the two variances can be used to test for the equality of variances. The ability of the Goldfeld-Quandt test to detect the heteroscedasticity problem is likely to be sensitive to the size of middle part, c, that should be discarded. In this work a simulation study was conducted to determine the appropriate value of c to make the Goldfeld-Quandt test more effective. The results of the simulation study confirmed that the appropriate size of the omitted values, c, should not be less than 30% of the sample size, n, in order to ensure a best performance of the Goldfeld-Quandt test.