Estimation, Evaluation, and Selection of Actuarial Models
Estimation, Evaluation, and Selection of Actuarial Models
Estimation, Evaluation, and Selection of Actuarial Models
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
105<br />
Exercise 25 The equations to solve are<br />
µ <br />
ln 18.25 − µ<br />
0.2 = F (18.25) = Φ<br />
σ<br />
µ <br />
ln 35.8 − µ<br />
0.8 = F (35.8) = Φ<br />
.<br />
σ<br />
The 20th <strong>and</strong> 80th percentiles <strong>of</strong> the normal distribution are −0.842 <strong>and</strong> 0.842 respectively. The<br />
equations become<br />
−0.842 = 2.904 − µ<br />
σ<br />
0.842 = 3.578 − µ .<br />
σ<br />
Dividing the first equation by the second yields<br />
−1 = 2.904 − µ<br />
3.578 − µ .<br />
The solution is ˆµ =3.241 <strong>and</strong> substituting in either equation yields ˆσ =0.4. The probability <strong>of</strong><br />
exceeding 30 is<br />
Pr(X >30) =<br />
µ <br />
ln 30 − 3.241<br />
1 − F (30) = 1 − Φ<br />
=1− Φ(0.4)<br />
0.4<br />
= 1− Φ(0.4) = 1 − 0.6554 = 0.3446.<br />
Exercise 26 For a mixture, the mean <strong>and</strong> second moment are a combination <strong>of</strong> the individual<br />
moments. The first two moments are<br />
E(X) = p(1) + (1 − p)(10) = 10 − 9p<br />
E(X 2 ) = p(2) + (1 − p)(200) = 200 − 198p<br />
Var(X) = 200 − 198p − (10 − 9p) 2 = 100 − 18p − 81p 2 =4.<br />
The only positive root <strong>of</strong> the quadratic equation is ˆp =0.983.<br />
Exercise 27 For the inverse exponential distribution,<br />
l(θ) =<br />
nX<br />
(ln θ − θx −1<br />
j<br />
− 2lnx j )=n ln θ − nyθ − 2<br />
j=1<br />
l 0 (θ) = nθ −1 − ny, ˆθ = y −1 ,wherey = 1 n<br />
nX<br />
j=1<br />
1<br />
x j<br />
.<br />
nX<br />
ln x j<br />
For Data Set B, we have ˆθ =197.72 <strong>and</strong> the loglikelihood value is −159.78. Because the mean<br />
does not exist for the inverse exponential distribution, there is no traditional method <strong>of</strong> moments<br />
estimate available. However, it is possible to obtain a method <strong>of</strong> moments estimate using the<br />
j=1