Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/62558
Title: | 門檻分量迴歸模型之分析 Essays in Threshold Quantile Regression |
Authors: | Christos Michalopoulos 米克里斯多斯 |
Advisor: | 管中閔(Chung-Ming Kuan) |
Keyword: | quantile regression,threshold regression,single and multiple thresholds,nonlinearity test,2-parameter Gaussian process,Brownian bridge,Wald statistic,Likelihood-Ratio test,shrinking and fixed shift asymptotics, |
Publication Year : | 2013 |
Degree: | 博士 |
Abstract: | This dissertation deals with estimation and inference of threshold quantile regression
models with one and multiple threshold values (change-points). On chapter one, we introduce the quantile regression estimation method and a nonlinear modeling approach called “threshold regression” and we motivate why it is a good idea to blend these approaches together in answering real economic data problems. On chapter 2, we formulate a threshold quantile regression model for one, known or unknown threshold value. We derive the asymptotic properties of the model parameters as well as the threshold value and develop inferential procedures to identify heterogeneous effects of different covariate quantile ranges on different quantiles of the response. We conduct inference by developing a sup-Wald test that converges to a two-parameter Gaussian process that generalizes that of Galvao et al. (2011) in allowing for serially correlated errors. In addition, we derive the limiting distribution of the estimated threshold value assuming asymptotically shrinking shifts and construct confidence intervals for the estimated threshold value via a Likelihood-ratio-type statistic. Simulation studies assess favorably our proposed methods. On chapter 3, we extend the modeling framework above to a multiple threshold quantile regression model with known or unknown threshold values and analyze the properties of the parameter estimators together with the estimated threshold values. We derive the limiting distribution of the threshold values under the asymptotic frameworks of “fixed” and “shrinking” magnitude of shifts and we discuss the case where threshold e ects on one quantile a ect neighboring quantiles as well. We develop a sup-Wald test to identify heterogeneous effects of different covariate quantile ranges on quantiles of the response and we propose a Likelihood-Ratio-type test for l versus l + 1 regime-changes in the covariate and derive its limiting distribution. Simulations assess favorably the relevance of our testing procedures. Our asymptotic results, complement and extend those of Galvao et al. (2011), Gonzalo and Pitarakis (2006) and Li and Ling (2011) to the quantile regression setting. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/62558 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 經濟學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-102-1.pdf Restricted Access | 685.46 kB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.