Speed check: copula selection and parameter estimation using non-parametric moment inverses and maximum likelihood
The purpose of this article is to conduct a Monte-Carlo simulation study that compares the efficiency, accuracy and empirical computation time of three different estimators in terms of selecting copulae and estimating their parameters. I will compare two moment-based estimators, the inverse of Kendall’s Tau and Blomqvist’s Beta, to the classic canonical maximum likelihood estimator. I find that, in terms of estimating parameters for a given copula, the moment-based estimators are inferior to maximum likelihood, especially for small sample sizes. For selecting copulae from a set of data, however, all three estimators are equally accurate while the moment-based ones are at least 16 times faster computationally.
The main article is fast_copula_estimation.pdf