home / centre for automotive safety research / Publications / List / Details Publication DetailsTitle  Discussion of "Importance of tail dependence in bivariate frequency analysis" by Annie Poulin, David Huard, AnneCatherine Favre, and Stephanie Pugin  Authors  Hutchinson TP  Year  2008  Type  Journal Article  Abstract  The copula is what defines the association between two random variables, distinct from their univariate distributions. It is a function (of two arguments), and thus is much more complicated than a correlation, which is a onenumber summary. Appreciation that the copula may be useful in hydrology dates back at least to the suggestion by Moran (1970, p. 109) that rainfall in seeded and nonseeded conditions might be distributed as a bivariate normal distribution transformed to gamma margins. That is, there are three questions: (a) what the first margin is, and in the case of seeded rainfall, the assumption might be that it is a gamma distribution as this is skew; (b) what the second margin is, and for rainfall in the nonseeded condition, the assumption might again be that it is a gamma distribution; (c) the nature of the association, and the assumption might be that the copula is that of a bivariate normal distribution if for no other reason than the familiarity of that distribution. (Moran 1970, p. 109 is not explicit concerning how the pairs of observations are obtained, but they might refer to two places randomized on one day such that one is seeded and the other is not.)  Journal Title  Journal of Hydrologic Engineering  Journal Volume (Issue)  13(12)  Page Range  1202 
Reference  Hutchinson TP (2008). Discussion of "Importance of tail dependence in bivariate frequency analysis" by Annie Poulin, David Huard, AnneCatherine Favre, and Stephanie Pugin. Journal of Hydrologic Engineering, 13(12), 1202. 
