The course schedule below will be filled in throughout the semester. For each day, there will be a link to course slides (if there are any that day) and notes/links on what is due ahead of class time. The “Announcements” column is (hopefully) self-explanatory. Any urgent announcements will be made over email.
1/20: Winter Break
Before the semester: (Optional!)
Read the course syllabus. I will assume that you understand all course policies beginning on Tuesday of next week! Familiarize yourself with Moodle and the course website Complete Introductions Survey Join Slack |
1/22: First Day of Class, Probability Review
Before class:
Read the course syllabus. I will assume that you understand all course policies beginning on Tuesday of next week! Familiarize yourself with Moodle and the course website Complete Introductions Survey Join Slack
During class:
|
Welcome (back) to campus!
- PS1 due Tuesday 2/10 (posted by Tuesday 1/27)
|
1/27: Frequentist Estimation
Before class:
Read through Section 1.3 of the Statistical Theory course notes, and make sure you’re familiar with all concepts other than Jacobians / Jacobian Matrices (we won’t use them in this course). Make sure you can answer the first two Concept Questions in Section 1.2 of the Statistical Theory course notes, and think about an answer to the third question. Make at least one slack post, either in the #discussion-boards channel related to the pre-class material, or in #random introducing yourself to people in class!
During class:
|
1/29: Bayesian Estimation
Before class:
Read through the beginning of Chapter 2 of the Statistical Theory course notes (you can skip the section “When maximizing the ‘usual’ way doesn’t work…”) Finish working through the solution to the MLEs for \(\mu\) and \(\sigma^2\) in the Normal model (worked example 2, Section 2.5). Check your answer with the solutions in the Statistical Theory course notes Make at least one slack post in #general (responding to someone else, with your own post, etc.) regarding the material we’ve covered so far (prob review + MLEs + anything on PS1 you’ve looked at). As a reminder, this can be a comment, question, or something “digging deeper” (what are you still thinking about, can you draw connections to other classes/topics/things you’ve read, etc.)
During class:
|
- PS1 due Tuesday 2/10
- Make an Overleaf account if you don’t already have one, and create a project for your 454 problem sets!
|
2/3: Bayesian Estimation II (conjugate priors)
Before class:
Read through Section 1.1 and 1.2 of Bayes’ Rules for some background on Bayesian thinking Read through Sections 1.1 - 1.3 of Bayesian Theory for another perspective on the history of Bayesian philosophy Remember, you need to average 2 slack posts each week! Here are a few options:
What stuck out to you in the history of Bayesian statistics? Is it a philosophy you align with? Why or why not? For the coded Beta(1,1)-Binomial example we did in class on Thursday, mess around with different n’s and x’s (maybe same p every time, or maybe same x every time and changing p) and visualize how your posterior changes. Post a summary plot of your findings and some thoughts in the #coding channel! Post something you’re stuck on. Do you have questions on the practice problems given at the end of class? Do you think you have an algebra error somewhere but can’t find it? Your classmates and I can help you de-bug your math!
During class:
Beta-Binomial Poisson-Gamma
Code:
|
2/5: Frequentist vs. Bayesian Estimators
Before class:
Read through Section 5.1 and 5.2 of Bayes’ Rules for some background on the conjugate priors we’ve covered so far Work through the algebraic details of the Normal-Normal conjugate model where the variance \((\sigma^2)\) is known (see equation 5.15 in Bayes’ Rules). Try it for yourself first, then check the solution in Section 5.3.4 of Bayes’ Rules. Slack post (as per usual). All the usual options apply! Here are three extra ones to post in #coding:
Plot the prior and posterior Beta distributions next to each other, like we’ve done before in class. Try changing prior parameters and likelihoods to see how things shift, and post your thoughts (and code) to the #coding channel Similar plotting question as above, but try it out with the Poisson-Gamma (this is a bit more challenging since I haven’t given you specific code for this one). If you are really code motivated, try writing a function to take in some prior parameters and data (likelihood) that produces a plot of the posterior, likelihood, and prior for a given conjugate family!
During class:
|
- PS1 due Tuesday 2/10
- Theory Exam: Tuesday 2/17
|
2/10: Metropolis Hastings
Before class:
Finish working through Q2b of the Monte Carlo activity we did in class last week (we’ll do Q2c - e in class today). If you’re less comfortable with some of the coding concepts we went over, make yourself a notesheet (maybe a Google Doc) with a list of useful functions and examples. Slack post (as per usual). All the usual options apply! Here are some extra options to post in #coding. If someone else already posted a response to one of these, you can post your response too, even if your findings are similar!!
In the Equality Index in the Monte Carlo activity, CA was a bit of an outlier. Repeat the steps of the analysis we went through with CA removed from the dataset, and post your updated findings and commentary (a plot, a summary, etc.) in #coding If you tried the Truel problem, post your code and findings! If you have any questions regarding any of the code we’ve been over so far, now would be a great time to ask!
During class:
- MH lecture & activity (see Recordings & Lecture tab!)
|
2/12: Metropolis Hastings & Gibbs Sampling
Before class:
REVIEW CODE! Get comfy with for-loops, if statements, and indexing! Slack post! You know the drill by now.
During class:
- Finish MH lecture & activity
- Gibbs lecture & activity (see Recordings & Lecture tab!)
|
- PS1 due Tuesday 2/10
- Theory Exam: Tuesday 2/17
- PS2 due Tuesday 3/3
|
2/17: Theory Exam
Before class: Study!!
During class: Theory Exam
|
2/19: Gibbs Practice
Before class:
Read through Sections 7.1-7.2 of Bayes Rules! (read 7.3 if you want to, it’s another way to code what we did in class) for a recap of the MH activity. Try rerunning the M-H algorithm from the example in class with half-width \(w = 0.05\) and \(w = 100\) (very small vs. very big). What do you notice in your traceplots, and your Markov chain samples relative to when we used \(w = 1\) in class? Check if your results / conclusions align with Section 7.4 of Bayes Rules! where they discuss “tuning” the M-H algorithm.
During class:
|
- Theory Exam: Tuesday 2/17
- Take a look at the Tech Report Pairings!
- Theory Exam Retakes: Schedule a 30-minute time slot with me with this link!
- PS2 due Tuesday 3/3
|
2/24: Stan
Before class:
Before reading the papers linked, take a look at a few suggestions about how to read an academic paper. You don’t need to actually take any of the suggestions if you have a system that works for you, but I find this one helpful, personally!
Please bring an annotated/marked-up copy of the first preprint linked below to class today! This doesn’t have to be on paper, iPad or computer is fine. This is just to help aid discussion so we can each reference our notes.
Read from the beginning through Section 3 of this preprint on Hamiltonian Monte Carlo for “physics dummies” (one of the more intuitive explanations of HMC)
- This preprint, Sections 1 through 3.4, offers another explanation of the high-level idea, if you’d like to read another person’s framing of the problem as well!
Check out this github page on getting started with rstan (the package that implements the HMC sampler we’ll be using!) Do not install the development version. Before class, try to work through the steps where you configure your C++ toolchain. We’ll work through the rest of the installation in class, but if you want to continue on your own beforehand, feel free! Complete the Gibbs activity from Thursday on your own outside of class for extra practice! There are plenty of things from the second example to comment on / code up for slack post fodder!
During class:
HMC discussion Stan install day |
2/26: Stan
Before class:
Have Stan installed! If you were unable to install in class on Tuesday, follow the steps here, including the “Verifying Installation” step. Read over the Technical Reports tab of the course website in detail. Come to class today with any questions you have about process, and make sure you’ve decided with your partner how you plan on communicating / working on the report together. Slack post! Either continue or contribute to a discussion about HMC, following up from our conversation in class on Tuesday, or ask a question about the problem set, or, or, or…
During class:
|
- PS2 due Tuesday 3/3
- Suggested half-way point is this Tuesday! Try to give yourself 1 week for the Gibbs problem, ideally.
- Take a look at the Tech Report Pairings!
- TR1 First Draft is due Tuesday 3/3
- Final Draft is due Monday 3/9 at 8:00am on Moodle
- Sign up with your partner for a time slot for your TR1 Oral Assessment! You should sign up during your usual class time if possible.
- Theory Exam Retakes: Schedule a 30-minute time slot with me with this link!
- MSCS Capstone days are 3/5 and 3/6! A list of talks can be found here (must be signed in to your Mac email to access)
|
3/3: TR1 Peer Review
Before class:
Make sure you have PS2 submitted on Moodle by 8am, unless you’re taking an extension! Bring two printed copies of your TR draft without your names (anonymized!) to class. I will number them as you come in to keep track of who is who. Make a shared Overleaf project with your tech report teammate, if you haven’t already, for collaboration purposes. This is where you should upload the editor comments and author comments templates to work on during class today. This is what you’ll email to me (Taylor) after class to submit your “Referee Report.”
During class:
|
3/5: MSCS Capstone Days
Class is cancelled for MSCS Capstone Days! You must attend a minimum of 3 capstone talks during MSCS Capstone days (March 5th and 6th). These can be during our usual class time or any other time you are free and interested in attending a talk!
If you are presenting during Capstone days, you may count your own talk as one of the 3 required reflections for this course.
The Capstone days schedule can be found here (must be signed in to your Mac email to access).
Before class next Tuesday:
|
- PS2 due Tuesday 3/3
- TR1 First Draft due Tuesday 3/3
- Final Draft is due Monday 3/9 at 8:00am on Moodle
- Sign up with your partner for a time slot for your TR1 Oral Assessment! You should sign up during your usual class time if possible.
- MSCS Capstone days are 3/5 and 3/6! A list of talks can be found here (must be signed in to your Mac email to access)
- Capstone topic selection (options & guidelines) document posted! You will asked to fill out a survey about your capstone project topic choice by Thursday of next week
|
3/10: TR1 Oral Assessment
Before class:
- Prep with your partner for the Oral Assessment (see detailed information on the Tech Report page of the website)
If you are looking for slack post suggestions, I would recommend posting thoughts you have about any of the capstone topics, or trying to gauge whether others in your section may be interested in similar applied topics as you! You could consider reading ahead in this section of Bayes Rules! to get a better understanding of what a hierarchical model is (we’ll formalize after Spring Break) if you plan on doing an applied project. Other options include continuing with the Intro Stan activity on your own, and posting about the code. Try to get through the #Stan Output section before class on Thursday!
During class:
NOTE: NO regularly scheduled class! Instead, you’ll attend your 15-minute oral assessment. |
3/12: Stan II
Before class:
During class:
- Intro Stan activity continued
|
- TR1 Final Draft due Monday 3/9 at 8:00am on Moodle
- Sign up with your partner for a time slot for your TR1 Oral Assessment! You should sign up during your usual class time if possible.
- Capstone topic selection (options & guidelines) document posted! You will be asked to fill out a survey about your capstone project topic choice by Thursday of next week
- PS3 due Tuesday 4/7 (one question posted that you can already do!)
- MSCS (Stats) Seminar on Wednesday at 4:40pm in OLRI 254!
|
Before class:
Finish the Intro Stan activity! Consider making a post on Slack about anything you found confusing, anything you found particularly helpful, or anything you still have questions on. Other slack post options, related to TR1 Findings:
Take a look at the TR that made the data “moot” by making the prior agree with it! (located in Recordings/Documents > Assorted Extra Notes) Comment your thoughts on this approach / write-up Run the example code related to the “return 0” issue that came up in a few TRs, and discuss your findings / thoughts (located in Recordings/Documents > Assorted Extra Notes)
During class:
- Hierarchical Models lecture + activity
|
Before class:
Finish working through the hierarchical models activity Read Chapter 15 of Bayes Rules! for an overview of what we covered in class on Tuesday OPTIONAL: Read Chapters 17.1 - 17.3 of Bayes Rules! for a bit of background on our specific problem we worked on (using different code). This may be particularly interesting if you’ve taken Correlated Data (random intercepts and random slopes mentioned)! Read Vehtari et al. (2021) Sections 1 - 3 (pages 1 - 11) to discuss in class today. Pay particular attention to \(\hat{R}\) (classic and split!) and ESS. Don’t forget the suggestions on how to read an academic paper!
During class:
- Convergence diagnostics discussion + lecture
|
|