统计学相关讲座

Double Asymptotics for Explosive Continuous Time Models个人简介

被阅览数：次  发布时间:2014/04/18 16:25:28

主讲人： Prof. Xiaohu Wang
主讲人简介：

 Chinese University of Hong Kong
简介：
This paper provides "double asymptotic" distributions of least squares (LS) estimators of the persistence parameter in univariate explosive continuous time models driven by Lévy processes. We also derive the "double asymptotic" distributions of the LS estimators of the autoregressive parameter in the correspondingly exact discrete-time model. The "double asymptotic" limit theory assumes the sample size diverges because (i) the sampling interval shrinks to zero and (ii) the timespan of the data diverges. Results are derived allowing for (i) and (ii) to occur simultaneously or sequentially. The intercept term and the initial condition are found to play significant roles in the limit distribution. The limit distribution can be used to find statistical evidence for explosive behavior in the continuous time set-up. The results in the paper contribute to the literature in three aspects. First, in comparison with the limit distribution given in Phillips and Magdalinos (2007), the "double asymptotic" distribution of the LS autoregressive parameter estimate provides an improved approximation to the finite sample distribution of the estimate by allowing it to rely on the intercept term and initial value of the process. Second, an approach is proposed to estimate and conduct inference on the continuous time persistence parameter. Third, we bridge the gap between the simultaneous double asymptotics and the sequential asymptotics given in Anderson (1959) and Perron (1991) by showing that all the limit distributions share the same form.
时间： 16:30-18:00, Wednesday, April 23, 2014
地点： Room N302, Economics Building
期数： 厦门大学统计学高级系列讲座
主办单位：
类型： 系列讲座

上一条：A Bayesian approach for identification and assessment of biomarkers predictive of cancer treatment个人简介
下一条：Estimating Social Intercorrelation with Sampled Network Data个人简介