douglaslam

July 24th 2006, 09:25 AM

I am not quite clear why cov(Y,Z)=-n[1-S(cj-1)][S(cj-1)-S(cj)], where Y is the number of observations at or below cj-1 and Z is the number of observations above cj-1 and at or below cj. This appears on p. 307 of Loss Models 2nd edition.

From the definition, cov(Y,Z) should be -E[Y]E[Z], so shouldn't n be replaced by n squared??????

From the definition, cov(Y,Z) should be -E[Y]E[Z], so shouldn't n be replaced by n squared??????