In the notes, it mentions that one of the benefits of rank correlation over linear correlation is that rank correlation is not dependent on the marginal distributions of the variables but linear correlation is. I'm struggling to understand why this is an advantage/disadvantage - can anyone help please? Thanks in advance Clare
Hi Clare Yes - an important question as it gets at how copulas might be fitted (i.e. parameterised). When you get to section 5.4 and the appendix of Module 18 you will see that there is a relationship between rank correlation coefficients and the parameters of single-parameter Archimedean copulas. It is this relationship that facilitates the fitting of such copulas to a particular dataset. (No such relationship exists for Pearson's correlation coefficient, due to the influence of the marginal distributions.) Best wishes David
