So there are many different credibility formulae in exam 7, pulled together from different actuarial papers. Are some of them meant to be equivalent? Please somebody help me reconcile before I explode.
1. Hurliman Z=p/(p+t) is supposably the optimal Z* that minimizes reserve MSE.
2. Brosius LS credibility Z=bp, which should also be optimal as it's LS. Should it equal Hurliman?
3. Brosium Baysian Z=VHM/(VHM+EVPV). There was a similar thing in exam 5 also. How doea it compare to the "optimal" Z's above?
4. Verrall Z=sum(y)/(beta*phi+sum(y)), which looks somewhay similar to Hurliman... and is the part beta*phi equivalent to Hurliman's t??
5. Mack (2000) commented that benktander is better tham BF if Z*>p/2. Substitute in Hurliman's formula I get p+sqrt(p)<2, which is almost always true. Does that mean benktander is almost always better than BF?
6. There's also a Neuhaus credibility... why why why?
Counting down to exam day! Good luck my friends.
1. Hurliman Z=p/(p+t) is supposably the optimal Z* that minimizes reserve MSE.
2. Brosius LS credibility Z=bp, which should also be optimal as it's LS. Should it equal Hurliman?
3. Brosium Baysian Z=VHM/(VHM+EVPV). There was a similar thing in exam 5 also. How doea it compare to the "optimal" Z's above?
4. Verrall Z=sum(y)/(beta*phi+sum(y)), which looks somewhay similar to Hurliman... and is the part beta*phi equivalent to Hurliman's t??
5. Mack (2000) commented that benktander is better tham BF if Z*>p/2. Substitute in Hurliman's formula I get p+sqrt(p)<2, which is almost always true. Does that mean benktander is almost always better than BF?
6. There's also a Neuhaus credibility... why why why?
Counting down to exam day! Good luck my friends.
Those gazilions of credibility formulae
0 commentaires:
Enregistrer un commentaire