Exam MAS-I Practice Question 102
Study guide: Exam MAS-I
Two Poisson GLMs with a log link are fitted to the same 3,000 records of claim counts, with the same rating variables:
| Model 1 | Model 2 | |
|---|---|---|
How \ln(\text{car-years}) enters |
offset | predictor |
Coefficient of \ln(\text{car-years}) |
1 (fixed) | 0.960 (SE 0.030) |
| Residual deviance | 3,412.6 | 3,410.8 |
| Residual degrees of freedom | 2,985 | 2,984 |
Which recommendation is best supported, for the reason given?
- A) Model 2, because its residual deviance is lower, so it fits the data better
- B) Model 2, because its coefficient of 0.960 shows claims rise less than in proportion to exposure
- C) Model 1, because the data are consistent with a coefficient of 1, so fixing it costs nothing
- D) Model 2, because the coefficient's
zvalue of 32 shows that exposure is highly significant - E) Model 1, because a model with an offset is not nested in the model without it, so no test applies
Answer this question and see the worked solution