Bootstrap Standard Error Estimates in a Switching Regression Model with Unknown Switch Point

Citation

Douglas, Stratford & Guilkey, David K. (1996). Bootstrap Standard Error Estimates in a Switching Regression Model with Unknown Switch Point. Communications in Statistics: Simulation and Computation, 25(1), 1-23.

Abstract

We identify and analyze a problem in switching regressions estimation, and evaluate the bootstrap as a solution. The usual estimation method for this model overstates the precision of the parameter estimates, since the standard error estimates that it generates are conditional on the switch point estimate. Bootstrap techniques allow the estimation of unconditional standard errors. We estimate a response surface to evaluate the accuracy and responsiveness of the bootstrap standard errors. Our results indicate that the bootstrap estimates standard errors of both the switch point estimator and regression coefficient estimators with reasonable accuracy.

URL

https://doi.org/10.1080/03610919608813295

Reference Type

Journal Article

Year Published

1996

Journal Title

Communications in Statistics: Simulation and Computation

Author(s)

Douglas, Stratford
Guilkey, David K.