The use of reduced variables eliminates the above nonhomogeneous term as it

# The use of reduced variables eliminates the above

• 10

This preview shows page 7 - 10 out of 10 pages.

The use of reduced variables eliminates the above non–homogeneous term as it multiplies a 0–expectation term. Using the fact that a scalar random variable trivially equals its trace and also that Tr( A B ) = Tr( B A ) we finally have E braceleftbig ( X 0 j ) prime c prime 1 Σ j bracketleftbig β 0 j ) - Z j ˆ β 0 j bracketrightbigbracerightbig = E braceleftbig Tr bracketleftbig ( X 0 j ) prime c prime 1 Σ j ( β 0 j ) - Z j ˆ β 0 j ) bracketrightbigbracerightbig = E braceleftbig Tr bracketleftbig c prime 1 Σ j ( β 0 j ) - Z j ˆ β 0 j ) ( X 0 j ) prime bracketrightbigbracerightbig = Tr bracketleftbig c prime 1 Σ j E braceleftbig( β 0 j ) - Z j ˆ β 0 j ) ( X 0 j ) prime bracerightbigbracketrightbig . This last expression is seen to be equal to 0 for Z j given in (6.5), since: E braceleftbig( β 0 j ) - Z j ˆ β 0 j ) ( X 0 j ) prime bracerightbig = E bracketleftbig β 0 j ) ( X 0 j ) prime bracketrightbig - Z j E bracketleftbig ˆ β 0 j ( X 0 j ) prime bracketrightbig = Cov bracketleftbig β 0 j ) , X 0 j bracketrightbig - Z j Cov[ ˆ β 0 j , X 0 j ] = Cov[ β j ) , X j ] - Z j Cov[ ˆ β j , X j ] = A Y prime - Z j ( A + s 2 U j ) Y prime = 0 . a50

Subscribe to view the full document.

66 CHAPTER 6. CREDIBILITY REGRESSION MODELS Lemma 6.3. To estimate b and Z j = A ( A + s 2 U j ) - 1 in (6.5), use ˆ s 2 = 1 k 1 ( n - p ) k summationdisplay j =1 [ X j - Y ˆ β j bracketrightbig prime V - 1 j [ X j - Y ˆ β j ] , ˆ A = 1 ( k - 1) k summationdisplay j =1 Z j [ ˆ β j - ˆ b ] [ ˆ β j - ˆ b ] prime , where ˆ b = ( k summationdisplay j =1 Z j ) - 1 k summationdisplay j =1 Z j ˆ β j , which are unbiased pseudo–estimators of the parameters s 2 , A and b , respec- tively. *Proof: Left as an exercise. Theorem 5.2.4 on page 183 of Goovaerts et al. (1990) gives a proof of the unbiasedness of ˆ b . Other suggested estimators as well formulas for unequal numbers of periods, n j , are also given. a50 6.3 A Numerical Illustration We now compare the estimates obtained in the previous chapters for Hachemeis- ter’s data set to the premiums obtained with a regression credibility model. We use Hachemeister’s standard design matrix Y = 1 12 1 11 . . . . . . 1 1 and a diagonal weight matrix V - 1 j = diag( w j 1 , . . . , w j 12 ). Other design matri- ces could be used, but in this way the predicted value of next years’ credibility premium, that is for t = 0 is simply obtained from the regression intercept and there is no need to pre–multiply by any vector of y j elements. These premiums were obtained with initial diagonal matrices Z j , equal to the identity, which, together with ˆ s 2 , produced the first iteration estimates of b . With it the first estimate of A was obtained, yielding a second estimated value of Z j . The stopping rule for this iterative procedure was that the
6.3. A NUMERICAL ILLUSTRATION 67 Table 6.1: Hachemeister’s Regression Premiums Individual Regression Estimators Class j = 1 2 3 4 5 Intercept 2,470 1,621 2,096 1,538 1,676 Slope -62 -17 -43 -28 -12 Credibility Adjusted Regression Estimators Intercept 2,437 1,651 2,073 1,507 1,759 Slope -57 -21 -41 -15 -26 largest difference between two successive estimates, of either the intercept or the slope, be at most 0 . 0001. Convergence occurred after 36 iterations.

Subscribe to view the full document.

• Fall '09
• Dr.D.Dryanov

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern

Ask Expert Tutors You can ask 0 bonus questions You can ask 0 questions (0 expire soon) You can ask 0 questions (will expire )
Answers in as fast as 15 minutes