Free CAS MAS-I (Modern Actuarial Statistics I) Extended Linear Models Practice Questions

Extended linear models on CAS Exam MAS-I cover generalized linear models (GLMs), link functions (log, logit, identity), model selection criteria, residual diagnostics, and applied regression for insurance ratemaking and classification (CAS).

321 questions 41 easy 171 medium 109 hard 2026 syllabus

Sample Questions

Question 1 Easy
Which of the following statements is TRUE regarding the use of an offset variable in a generalized linear model?
Solution
C is correct. An offset is a known quantity that enters the linear predictor with a coefficient fixed at 1 (rather than being estimated from the data). For a Poisson GLM with log link, the linear predictor is η=β0+β1x1+…+log⁡(exposure)\eta = \beta_0 + \beta_1 x_1 + \ldots + \log(\text{exposure}), so the predicted count scales proportionally with exposure. This is the standard mechanism for handling unequal time-at-risk or other normalizing quantities in count and severity models.
Question 2 Medium
An actuary fitting a Poisson GLM to auto insurance claim count data observes that the Pearson chi-squared statistic divided by its degrees of freedom is 3.8.

Which of the following best describes this situation and the appropriate response?
Solution

(A) is correct.

For a well-fitting Poisson model, the dispersion parameter ϕ^=X2/(n−p)\hat{\phi} = X^2 / (n - p) should be close to 1, since the Poisson distribution has Var(Y)=μ\text{Var}(Y) = \mu (variance equals mean, i.e., ϕ=1\phi = 1). (C) ratio of 3.8 is substantially greater than 1, indicating overdispersion: the observed variance is roughly 3.8 times the variance assumed by the Poisson model. The standard remedies are quasi-Poisson (which rescales standard errors by ϕ^\sqrt{\hat{\phi}}) or negative binomial regression (which has an extra dispersion parameter).
Question 3 Hard
An actuary fits a linear regression model with 8 predictors to a dataset of n=200n = 200 homeowners insurance claims. The model has R2=0.64R^2 = 0.64 and adjusted R2=0.625R^2 = 0.625.

Which of the following best explains why adjusted R2R^2 is lower than R2R^2?
Solution
C is correct.

Adjusted R2R^2 is defined as:
Radj2=1−(1−R2)(n−1)n−p−1R^2_{\text{adj}} = 1 - \frac{(1 - R^2)(n - 1)}{n - p - 1}
where pp is the number of predictors. Because n−1>n−p−1n - 1 > n - p - 1 when p≥1p \geq 1, the fraction (n−1)/(n−p−1)(n-1)/(n-p-1) exceeds 1, which inflates the penalty term (1−R2)(1-R^2). This means adjusted R2≤R2R^2 \leq R^2, with equality only when p=0p = 0. The adjustment penalizes adding predictors that do not sufficiently improve fit, discouraging overfitting.

Here: Radj2=1−(1−0.64)(200−1)200−8−1=1−0.36×199191=1−71.64191=1−0.375=0.625R^2_{\text{adj}} = 1 - \frac{(1 - 0.64)(200 - 1)}{200 - 8 - 1} = 1 - \frac{0.36 \times 199}{191} = 1 - \frac{71.64}{191} = 1 - 0.375 = 0.625, confirming the stated value.

Guides & Articles

About FreeFellow

Jeffrey Ting, founder of FreeFellow
Jeffrey Ting
FSA, CFA · Founder

FreeFellow was built by Jeffrey Ting, a credentialed actuary and CFA charterholder who has passed thirteen of the hardest exams in finance, all on his first attempt. He paid four-figure prep fees along the way.

So he started writing his own questions, then lessons, then mock exams, until it grew into a full prep platform covering 41 exams with more than 48,000 original practice questions.

01
Cost shouldn't decide who gets in.

A CFA charter, a CPA license, an actuarial credential. Each opens a real career, but prep fees add to the cost of getting there. FreeFellow keeps its original question bank and written lessons free so you can study even if a paid course is out of reach.

02
Free should mean free.

The original question bank, the worked solutions, the topic lessons, mixed practice, and your readiness score stay free, forever. Full access needs a free account, and practice usage limits apply. Fellow adds AI grading on supported exams, spaced-repetition flashcards, full-length mock exams, analytics, and a study plan that adapts to your progress.