multibreak |

Multiple breakpoint testing.

The multibreak view of an equation displays the results of multiple breakpoint testing using sequential and global optimization methods.

This view is only available for (non-panel) equations specified by list without ARMA terms and estimated by ordinary least squares.

Syntax

equation_name.multibreak(options) [list_of_breaking_regressors]

where equation_name is the name on an equation specified by list and estimated using least squares. The multibreak may be followed by options, and an optional list of breaking regressor names. If the latter list is omitted, the coefficients for all of the regressors in the original equation will be allowed to vary across regimes.

Options

method=arg (default=“seqplus1”) | Breakpoint testing method: “seqplus1” (sequential tests of single versus breaks), “seqall” (sequential test of all possible versus breaks), “glob” (tests of global vs. no breaks), “globplus1” (tests of versus globally determined breaks), “globinfo” (information criteria evaluation). |

trim=arg (default=5) | Trimming percentage for determining minimum segment size (5, 10, 15, 20, 25). |

maxbreaks=integer (default=5) | Maximum number of breakpoints to allow (not applicable if “method=seqall”). |

maxlevels=integer (default=5) | Maximum number of break levels to consider in sequential testing (applicable when “method=sequall”). |

size=arg (default=5) | Test sizes for use in sequential determination and final test evaluation (10, 5, 2.5, 1) corresponding to 0.10, 0.05, 0.025, 0.01, respectively |

heterr | Assume regimes specific error distributions in variance computation. |

commondata | Assume a common distribution for the data across segments (only applicable if original equation is estimated with a robust covariance method, “heterr” is not specified). |

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

p | Print the view. |

Examples

equation eq01.ls m1 c tb3 gdp

eq01.multibreak(maxbreaks=3)

eq01.multibreak(method=glob, size=10, trim=15) tb3

The first test line tests for up to 3 structural breaks in all of the coefficients using sequential tests of single versus breaks. The second line tests uses the global breaks versus none tests with trimming value 0.15, and a size of 0.10 to test for differences in the coefficient on TB3 across regimes.

The multiple breakpoint tests will use the covariance matrix settings from the original equation when constructing tests. The command

equation eq01.ls(cov=hac, covkern=quadspec, covlag=1, covbw=andrews) rates c

eq01.multibreak(heterr)

eq01.multibreak(method=glob, heterr)

eq01.multibreak(method=globinfo)

estimate an equation using HAC covariances. The second line tests for up to 5 structural breaks using sequential tests of single versus breaks. The third line uses the global breaks versus none tests. Both of these tests allow for error distributions to vary across the different segments. The final line evaluates the breakpoints using information criteria associated with the global optimizers.

Cross-references

See “Multiple Breakpoint Tests” for discussion. See also “Least Squares with Breakpoints” for tools which estimate equations with structural breaks.

See Equation::breakls for estimation of regression equations with breaks.