Copula Specification Tests under General Censorship

We propose a family of data-driven tests for the goodness-of-fit of copula-based multivariate survival models under general censorship. Appealing features of the tests include flexibility, ease of implementation, distribution free asymptotic distributions and informativeness regarding alternative copulas when a null distribution is rejected. Consistency and large sample properties of the tests, with parametrically or nonpara-metrically estimated marginal distributions, are established. Monte Carlo simulations demonstrate good finite sample performance of the proposed tests. The semiparamet-ric tests are shown to rival the correctly specified parametric tests and at the same time are immune from risk of misspecification. Two empirical applications illustrate the usefulness of the proposed tests