Semiparametric estimation and model selection for conditional mixture copula models

Abstract

Conditional copula models allow the dependence structure among variables to vary with covariates, and thus can describe the evolution of the dependence structure with those factors. This paper proposes a conditional mixture copula which is a weighted average of several individual conditional copulas. We allow both the weights and copula parameters to vary with a covariate so that the conditional mixture copula offers additional flexibility and accuracy in describing the dependence structure. We propose a two-step semiparametric estimation method and develop asymptotic properties of the estimators. Moreover, we introduce model selection procedures to select the component copulas of the conditional mixture copula model. Simulation results suggest that the proposed procedures have a good performance in estimating and selecting conditional mixture copulas with different model specifications. The proposed model is then applied to investigate how the dependence structures among international equity markets evolve with the volatility in the exchange rate markets.

The Impact on Higher Education under COVID-19 Pandemic Prevention and Control in China

“…Recently, in order to encourage the development of educational information technology and the sharing and balancing of top-notch educational resources, Massive open online courses (MOOCs) have been developed with the help of educational authorities, with the active involvement of universities, and are available to the entire nation for free [1]. However, there is a serious issue with MOOCs being "reconstructed but not reused."…”

Copula-based Modeling for IBNR Claim Loss Reserving

“…This is while that the classical correlation measures such as Pearson's correlation coefficient only measures linear associations between marginal distributions. There are many studies to discuss how to select a copula for a given dataset, see [13,39]. The Clayton copula is an asymmetric Archimedean copula, which is able to measure positive dependency between random variables.…”

Gaussian Mixture Models with Concave Penalized Fusion

“…Finite mixture models are often used to model heterogeneous data from complex distributions, which find applications in many areas such as density estimation (Escobar and West, 1995), pattern clustering (Liu et al, 2022), and quality control (Li et al, 2021). As the most popular mixture model, the Gaussian mixture model (GMM) possesses many appealing features including computational tractability, affine invariance, and flexibility of representations.…”