GENERALIZED LAPLACE INFERENCE IN MULTIPLE CHANGE-POINTS MODELS

Under the classical long-span asymptotic framework, we develop a class of generalized laplace (GL) inference methods for change-point dates in linear time series regression model with multiple structural changes analyzed in, e.g., Bai and Perron (1998, Econometrica 66, 47–78). The GL estimator is defined by an integration rather than optimization-based method relies on LS criterion function. It interpreted as (non-Bayesian) estimator, proposed retain frequentist interpretation. This approach provides better approximation about uncertainty data change-points relative to existing methods. On theoretical side, depending some input (smoothing) parameter, estimators exhibits dual limiting distribution, namely shrinkage distribution or Bayes-type distribution. We propose based highest density regions using latter show that it has attractive properties not shared other popular alternatives, i.e., bet-proof. Simulations confirm these translate good finite-sample performance.