How did you answer the confidence interval (3 pointer) problem when they
talk about the variance of the estimator of exp(-x/lambda) ? I computed
the confidence interval as 1.96*std dev and got ans (C). What I should
have done was to take 2*1.96*std dev and got the answer E.
-------------
Original Text
From: <Joseph_C._Hebert@tiginsurance.com>, on 5/6/99 12:09 PM:
How did you interpret question 18.
The info for problem 18 and 19 stated that 1000 claims are required for
full credibility for claim frequency. Problem 18 asked for the number of
problems required for 50% credibility as stated in Hossak, et al.
So I assumed that they want 50% credibility for aggregate claim costs, not
50% credibility for frequency only. The aggregate claim costs assumption
yields 1250 claims ((.50)^2 * (1000(1+2^2)) Answer E. But if they wanted
50% credibility for frequency only the answer would be ((.50)^2 *
(1000))=250 answer B.
I choose answer E since the typical calculation on the syllabus and on
prior exams asks for full credibilty for aggregate claim costs, and only
shows frequency only or severity only credibility as special cases.