June 11, 2013 NR’s Ingrid Hobæk Haff received the Norwegian Statistical Association’s Sverdrup award for young researchers 2013 for the article «Parameter estimation for pair-copula constructions», doi: 10.3150/12-BEJ413. Ingrid Hobæk Haff works as research scientist at Norsk Regnesentral, department SAMBA. The article is part of her PhD dissertation «Pair-copula constructions – an inferential perspective» and was published in the renowned journal Bernoulli.

The award can be given to members of the Norwegian Statistical Association younger than 40 years old who has distinguished themselves in statistical research. Read more about the award (in Norwegian).

Parameter estimation for pair-copula constructions

Multivariate models constitute an increasingly important field within statistics. They describe the covariation of several variables, for instance the values of the various assets of a financial portfolio or meteorological entities, such as temperature, precipitation and wind speed. Ingrid Hobæk Haff has studied a multivariate model model called pair-copula construction.

This model has a series of qualities; it can link variables that behave quite differently both individually and jointly, and it handles extreme events. The theoretical foundation, provided by Hobæk Haff, makes it possible and safe to use the model in practice, e.g. for risk management or flood prediction.

Using a financial portfolio as a starting point, it is easier to characterise each asset than modelling their sum directly. This requires a specification of both the individual and joint behaviour of the variables. In extreme cases, such as a market crash or a flood, the variables’ mutual interaction becomes particularly important. Traditional models, such as the multivariate normal distribution, tend to underestimate this covariation, which may induce grave errors. Pair-copula constructions are designed for such cases. They have already been applied within a number of fields, counting finance, insurance, hydrology, genetics and climate modelling. However they are relatively new, and so far, the necessary theory is incomplete. This work is an important contribution in that respect.