KOTORY
Tests for heteroscedasticity in the linear regression model that sort the data by a regressor, split them into three equal parts and compare the error scale of the parts. The ordinary least squares version 'kah3.test()' refers the ratio of the largest to the smallest residual mean square to its exact null distribution, Hartley's maximum F-ratio with three groups; the robust version 'kah.robust.test()' replaces the mean squares by least trimmed squares scales, so that outliers neither create nor hide heteroscedasticity, and refers the ratio to a maximum F-ratio with simulated effective degrees of freedom, to a Monte Carlo reference or to a residual bootstrap. The distribution, density, quantile and random generation functions of the maximum F-ratio are provided, together with 'run.all.het()', which runs the proposed tests next to the Goldfeld-Quandt, Breusch-Pagan, White and robust modified Goldfeld-Quandt tests in one call. For more details see Hartley (1950) <doi:10.1093/biomet/37.3-4.308>, Goldfeld and Quandt (1965) <doi:10.1080/01621459.1965.10480811> and Rousseeuw (1984) <doi:10.1080/01621459.1984.10477105>.
This page was built for software: KOTORY
