Proxy methods for factor copula-based clustered data with covariates
Abstract
A proxy method is proposed for estimating parameters and latent factors for clustered data based on factor copulas, assuming a very large number of observations per cluster. The model can also incorporate covariates, and the conditional distribution of the response variable, considering the covariates, is a parametric family of continuous or discrete distributions, while the effect of a latent variable common to a cluster is modeled using a factor copula. Additionally, a goodness-of-fit test is proposed. This extends previous work on similar models where the number of observations per cluster was assumed to be small.