hettest |

Test for Heteroskedasticity.

Performs a test for heteroskedasticity among the residuals from an equation.

The test performed can be a Breusch-Pagan-Godfrey (the default option), Harvey, Glejser, ARCH or White style test.

Syntax

equation_name.hettest(options) variables

Options

type = keyword | where keyword is either “BPG” (Breusch-Pagan-Godfrey - default), “Harvey”, “Glejser”, “ARCH”, or “White”. |

c | include cross terms for White test. |

lags = int | set number of lags to use for ARCH test. (Only applies when type = “ARCH”. |

prompt | Force the dialog to appear from within a program. |

Variables

A list of series names to be included in the auxiliary regression. Not applicable for ARCH or White type tests. The following keywords may be included:

@regs | include every regressor from the original equation. |

@grads | include every gradient in the original equation (non-linear equations only). |

@grad(int) | include the int-th gradient. |

@white(key) | include white-style regressors (the cross-product of the regressor list, or the gradient list if non-linear). key may be on of the following keywords: “@regs” (include every regressor from the original equation), “@drop(variables)” (drop a variable from those already included. For example, “@white(@regs @drop(x2))” would include all original regressors apart from X2), “@comp” (include the compatible style White regressors, i.e. levels, squares, and cross-products). |

@arch(lag_structure) | include an ARCH specification with the number of lags specified by lag_structure. If lag_structure is a single number, then it defines the number of lags to include. Otherwise, the lag structure is in pairs. For example, “@arch(1 5 9 10)” will include lags 1, 2, 3, 4, 5, 9, 10. |

@uw(variables) | include unweighted variables (only applicable in a weighted original equation). |

Examples

eq1.hettest(type=harvey) @white(@regs @drop(log(ip)))

performs a heteroskedasticity test with an auxiliary regression of the log of squared residuals on the cross product of all the original equation’s variables, except LOG(IP).

Cross-references

See
“Heteroskedasticity Tests” for a discussion of heteroskedasticity testing in EViews.