Am I crazy, or is part a actually testing the assumption that incremental losses are proportional to cumulative losses to date? Not accident year independence. If you find that incremental ldfs are correlated, that suggested that expected incremental losses are dependent on not just cumulative losses to date but also the losses to date two ages prior (I.e. high development followed by low development). How does this violate accident year independence?
2014 #4 part b.
0 commentaires:
Enregistrer un commentaire