Scribd
Upload
42 views

Stat 532 Name: in Class Final 12/12/2016

This document appears to be an in-class final exam for a statistics course. The exam consists of questions about Bayesian modeling and MCMC algorithms. Specifically, it asks the student to: 1) Sketch out the Metropolis-Hastings algorithm and acceptance rate. 2) Sketch out the Gibbs sampler algorithm. 3) Fit a Poisson regression model using MCMC, explain model priors and proposal distribution, and how to make inferences and predictions. 4) Choose five of six additional questions to answer, covering topics like Jeffrey's priors, model checking, Bayesian vs. ridge regression, and MCMC acceptance rates.

Uploaded by

Tanjil Ahmed
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
42 views

Stat 532 Name: in Class Final 12/12/2016

This document appears to be an in-class final exam for a statistics course. The exam consists of questions about Bayesian modeling and MCMC algorithms. Specifically, it asks the student to: 1) Sketch out the Metropolis-Hastings algorithm and acceptance rate. 2) Sketch out the Gibbs sampler algorithm. 3) Fit a Poisson regression model using MCMC, explain model priors and proposal distribution, and how to make inferences and predictions. 4) Choose five of six additional questions to answer, covering topics like Jeffrey's priors, model checking, Bayesian vs. ridge regression, and MCMC acceptance rates.

Uploaded by

Tanjil Ahmed
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Stat 532 Name:

In class final
12/12/2016

1. (8 points) Sketch out the Metropolis-Hastings algorithm including the acceptance rate.

2. (7 points) Sketch out the algorithm for a Gibbs Sampler.


Stat 532 In class final - Page 2 of 4 12/12/2016

3. Assume you are asked to fit a Poisson regression model for y as a function of x1 , x2 , and the
intercept.
(a) (5 points) Write out the model, including link functions and all necessary priors.

(b) (3 points) Assume you plan to fit the model using a Metropolis-Hastings algorithm, ex-
plain your choice of the proposal distribution.

(c) (3 points) How will you make inferences about the parameters in the model?

(d) (3 points) How will you make predictions for a future set of X ∗ = [x1 , x2 ].
Stat 532 In class final - Page 3 of 4 12/12/2016

For questions 4-9 choose five of the six questions to answer.


4. (4 points) Why is a Jeffrey’s prior useful and how is it calculated?

5. (4 points) Describe your personal philosophy for choosing priors.

6. (4 points) Given a dataset, how do you verify that your assumed model and prior are reason-
able?
Stat 532 In class final - Page 4 of 4 12/12/2016

7. (4 points) Describe the similarities and differences between Bayesian regression and ridge re-
gression.

8. (4 points) Why do we track the acceptance rate in a Metropolis-Hastings algorithm and how
does it relate to effective mixing for posterior samples?

9. (4 points) Summarize a paper a classmate presented (not your own work).

You might also like