(a) 我们知道基准类别的系数都设置为统一。然后我们发现参考=exp(12 506405) exp(0 435276) exp(1 025997) exp(0 912526) exp(0 331272)exp( 12 506405) exp(0 435276) 1 1 1= exp(1 025997) exp(0 912526) exp(0 331272) = 9 68

(b) 由于交互假人，我们必须区分年轻男性和其他男性。对于年轻男性与年轻女性，比例为参考
=exp(12 506405) exp(1 025997) exp(0 912526) exp(0 331272)exp( 12 506405) 1 exp(0 912526)1
(1)= exp(1 025997) exp(0 331272) = 3 89

(2)因此，年轻男性保单持有人的预期索赔频率是其他人的 3.89 倍同等女性保单持有人。

(3)= exp(1 025997) = 2 79 (4)

(c) 没有一个模型是首选的。因为两个模型中的因变量不同（基于不同的数据集），比较 AIC 没有意义。所以我们不能说什么关于哪个模型在 AIC 方面是最优的。

Exercise 1
(a) We know that the coefﬁcients of the benchmark category are all set to unity. We then ﬁnd Rec f = exp( 12 506405)  exp(0 435276)  exp(1 025997)  exp(0 912526)  exp(0 331272) exp( 12 506405)  exp(0 435276)  1  1  1= exp(1 025997)  exp(0 912526)  exp(0 331272) = 9 68
Hence, the expected claim frequency of the given male policyholder is 9.68 times that of the given female policyholder.
(b) Because of interaction dummy, we have to make a distinction between young males and other males. For young males vs. young females, the ratio is Rec f = exp( 12 506405)  exp(1 025997)  exp(0 912526)  exp(0 331272) exp( 12 506405)  1  exp(0 912526)  1
(1)= exp(1 025997)  exp(0 331272) = 3 89

(2)Hence, the expected claim frequency of a young male policyholder is 3.89 times that of a ceteris paribus female policyholder.
For older males vs. older females, the ratio is Rec f=exp( 12 506405)  exp(1 025997)  1  1 exp( 12 506405)  1  1  1
(3)= exp(1 025997) = 2 79 (4)
Hence, the expected claim frequency of an older male policyholder is 2.79 times that of a ceteris paribus female policyholder.
(c) None of the models is preferred. Because the dependent variable in the two models in different
(based on different datasets), it does not make sense to compare the AICs. So we cannot say anything about which model is optimal in terms of AIC.