Venter Factors

Question 1

Formulas for adjusted SSE

Answer 1

SSE/(n - p)^2

Question 2

Formula for AIC

Answer 2

SSE*e^(2p/n)

Question 3

Formula for BIC

Answer 3

SSE*n^(p/n)

Question 4

Venter: Row times column parameter

Constant variance for all cells

Answer 4

f(d) = sum(h*q)/sum(h^2)
h(w) = sum(f*q)/sum(f^2)

Question 5

Venter: Modeled using f(d)h(w), assume variance proportional to incremental losses

Answer 5

f(d)^2 = sum(q/h)/sum(h)
h(w)^2 = sum(q/f)/sum(f)
BF type method

Question 6

Venter: Modeled using f(d)c(w,d), assume variance is constant for all cells

Answer 6

f(d) = sum(c*q)/sum(c^2)

CL method, OLS average

Question 7

Venter: Modeled using f(d)c(w,d), assume variance is proportional to reported losses

Answer 7

f(d) = sum(q)/sum(c)

CL method, volume weighted

Question 8

Venter: Modeled using f(d)c(w,d), assume variance is proportional to square of reported loss

Answer 8

f(d) = sum(q/c)/sum(1)???

Question 9

Venter: # of parameters in CL Method, m x m traingle

Answer 9

m - 1

Question 10

Venter: # of parameters in BF method, m x m triangle

Answer 10

2m - 2

Question 11

Venter: # of parameters in CC method, m x m triangle

Answer 11

m - 1

Question 12

Venter: Losses modeled as f(d)h(w), and we group row and column parameters: How many parameters?

Answer 12

2m - 1 (don’t count first row, but count first column)

Question 13

Test for linearity

Answer 13

Sequence of negative followed by positive residuals, may not be linear

Question 14

Test for Correlation of LDFs

Answer 14

T = r[(n - 2)/(1 - r^2)]^0.5
T is t-distributed with n-2 D.O.F.

If more than 0.1m + sqrt(m) are correlated, strongly suggested to be correlated; m = number of testable pairs = (n - 3 choose 2)

Question 15

Model inflation in BF method

Answer 15

q(w,d) = f(d)h(w)g(w+d)

g(w+d) is inflation term, could be (1 + j)^(w+d)

Question 16

Venter test for high or low diagonals

Answer 16

Regress losses against:
Cumulative losses at prior age
Dummy variable for each diagonal

Question 17

Making fewer parameters

Answer 17

CY affect becomes inflation, g(w+d) = (1+j)^(w+d)
AY effect, set h(w) = h(1+k)^w
age parameters, set f(d) = (1+i)^d

Model is h(1+i)^d(1+j)^(w+d)(1+k)^w [3 parameters]

Question 18

Venter: Cape Cod h

Answer 18

h = sum(f*q)/sum(f^2) over the entire triangle