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.
4.5. HYPOTHESIS TESTS 75<br />
4.5.2 Anderson-Darling test<br />
This test is similar to the Kolmogorov-Smirnov test, but uses a different measure <strong>of</strong> the difference<br />
between the two distribution functions. The test statistic is<br />
Z u<br />
A 2 [F n (x) − F ∗ (x)] 2<br />
= n<br />
t F ∗ (x)[1 − F ∗ (x)] f ∗ (x)dx.<br />
That is, it is a weighted average <strong>of</strong> the squared differences between the empirical <strong>and</strong> model distribution<br />
functions. Note that when x is close to t or to u, the weights might be very large due<br />
to the small value <strong>of</strong> one <strong>of</strong> the factors in the denominator. This test statistic tends to place more<br />
emphasis on good fit in the tails than in the middle <strong>of</strong> the distribution. Calculating with this<br />
formula appears to be challenging. However, for individual data (so this is another test that does<br />
not work for grouped data), the integral simplifies to<br />
kX<br />
A 2 = −nF ∗ (u)+n [1 − F n (y j )] 2 {ln[1 − F ∗ (y j )] − ln[1 − F ∗ (y j+1 )]}<br />
j=0<br />
kX<br />
+n F n (y j ) 2 [ln F ∗ (y j+1 ) − ln F ∗ (y j )]<br />
j=1<br />
where the unique non-censored data points are t = y 0