A0532
Title: Detection of threshold points in mean and variance forthreshold regression models
Authors: ChihHao Chang - National University of Kaohsiung (Taiwan) [presenting]
Abstract: The threshold regression model is considered with one threshold point in the mean and in the variance, respectively, for dependent data with heteroscedasticity. We denote the threshold regression model in the mean without the continuity constraint at the threshold point by M2. We then provide an ordered iterative least squares (OiLS) method when estimating M2 and establish the consistency of the OiLS estimator under mild conditions. We denote the model by M1 when the continuity constraint is imposed on the threshold regression model. Further, we denote the model with no threshold points by M0 and apply a model selection procedure to select the three models. We establish the selection consistency under regularity conditions. The same estimation and selection procedures are further applied to detect the threshold point in the variance of the models.
