Bounded-influence robust estimation of copulas

Samuel Orso (University of Geneva)
Monday, February 2, 2015 - 2:00pm
Spandauer Straße 1, Room 23

Copula functions are very convenient for modeling multivariate observations. Popular estimation methods are the maximum likelihood and a pseudo likelihood. Unfortunately, the resulting estimators can often be biased whenever relatively small model deviations occur at the marginal and copula levels. In this paper, we propose two robust estimators that do not share this undesirable feature. Since skewed and heavy tailed parametric marginals are considered in many applications, we also propose a computationally efficient robust estimator for such distributions that is corrected for consistency by means of indirect inference. We exhibit the performance of our robust estimators in a simulation study. We show insights of their applicability to income mobility of swiss households for the period 2011 and 2012 and to environmental data.