EVT and alternatives [16]

Go to: Summary | Previous | Next   
Bullet points include: QQ-plot of GPD is convex upwards (if xi>= 0). If data is Normal then fitting a GPD may overstate size of extreme events somewhat. Since Normal has same tail characteristics as xi = 0 GPD. In real life often have multiple series / loss types. Can construct multivariate EVT theory, but more complex. E.g. McNeil, Frey & Embrechts (2005)

Contents | Prev | Next | Library

Desktop view | Switch to Mobile