The paper considers joint maximum likelihood (ML) and semiparametric (SP)
estimation of copula parameters in a bivariate t-copula. Analytical expressions
for the asymptotic covariance matrix involving integrals over special functions
are derived, which can be evaluated numerically. These direct evaluations of
the Fisher information matrix are compared to Hessian evaluations based on nu-
merical di®erentiation in a simulation study showing a satisfactory performance
of the computationally less demanding Hessian evaluations. Individual asymp-
totic con¯dence intervals for the t-copula parameters and the corresponding tail
dependence coe±cient are derived. For two ¯nancial datasets these con¯dence
intervals are calculated using both direct evaluation of the Fisher information
and numerical evaluation of the Hessian matrix. These con¯dence intervals are
compared to parametric and nonparametric BCA bootstrap intervals based on
ML and SP estimation, respectively, showing a preference for asymptotic con¯-
dence intervals based on numerical Hessian evaluations.
«
The paper considers joint maximum likelihood (ML) and semiparametric (SP)
estimation of copula parameters in a bivariate t-copula. Analytical expressions
for the asymptotic covariance matrix involving integrals over special functions
are derived, which can be evaluated numerically. These direct evaluations of
the Fisher information matrix are compared to Hessian evaluations based on nu-
merical di®erentiation in a simulation study showing a satisfactory performance
of the computationally l...
»