A method of moments estimator of tail dependence in meta-elliptical models

作者：Krajina Andrea^*

来源：Journal of Statistical Planning and Inference, 2012, 142(7): 1811-1823.

DOI：10.1016/j.jspi.2012.01.020

• Asymptotic normality
• Elliptical copula
• Elliptical distribution
• Meta-elliptical model
• Method of moments
• Semi-parametric model
• Tail dependence

A meta-elliptical model is a distribution function whose copula is that of an elliptical distribution. The tail dependence function in such a bivariate model has a parametric representation with two parameters: a tail parameter and a correlation parameter. The correlation parameter can be estimated by robust methods based on the whole sample. Using the estimated correlation parameter as plug-in estimator, we then estimate the tail parameter applying a modification of the method of moments approach proposed in the paper by Einmahl et al. (2008). We show that such an estimator is consistent and asymptotically normal. Further, we derive the joint limit distribution of the estimators of the two parameters. We illustrate the small sample behavior of the estimator of the tail parameter by a simulation study and on real data, and we compare its performance to that of the competitive estimators.

• 出版日期2012-7