March 20th, 2015, 11:55 am
Thanks Alan.Tagoma (and Alan and everybody): I'm studying a model which envolves the computation of higher order moments of underlying returns as well as the Breeden Litzenberger formula (ie getting state-prices from option prices).Now, in order to compute higher order moments with sufficient statistical significance I need a sufficiently long time-series of returns. However, given the nature of the model, for each data point I also need options. So I need sufficiently long time-series of option prices too.My conclusion is that for highly non-gaussian distributions we need so many data that the above model is likely not implementable. Or eventually it is but only for few assets for which we have liquid data for a wide range of strikes and for a long time-series.Any comment?