I’m soooo happy we’ll be going back to regular school at Columbia this fall. Those remote classes were getting worse and worse, and I’m really looking forward to being able to walk around the classroom as the students are working together in pairs. I bought this remote mic-and-speaker thing so I can speak without shouting while wearing a mask in class and the students will still be able to hear me.
I’ll be teaching 2 classes this semester:
POLS 4710: This is the first semester of a two-semester course on applied regression and causal inference based on our book, Regression and Other Stories. The two-semester sequence is for political science M.A. students, but anyone can take either the first semester (POLS 4710), the second semester (4712), or both classes. The fall class covers chapters 1-12 of the book. Topics covered in POLS 4710 include:
• Applied regression: measurement, data visualization, modeling and inference, transformations, and linear regression.
• Simulation, model fitting, and programming in R.
• Key statistical problems include adjusting for differences between sample and population, adjusting for differences between treatment and control groups, extrapolating from past to future, and using observed data to learn about latent constructs of interest.
• We focus on social science applications, including but not limited to: public opinion and voting, economic and social behavior, and policy analysis.
The course is set up using the principles of active learning, with class time devoted to student-participation activities, computer demonstrations, drills, and discussion problems.
The primary audience for this course is Poli Sci M.A. students as well as less quantitatively-inclined Poli Sci Ph.D. students. POLS 4710 should be ideal for social science graduate students in other departments and schools as well.
Here’s the syllabus for POLS 4710 this fall.
STAT 6106: This is a one-semester course on communicating data and statistics, covering the following topics:
1. Introducing yourself and telling a story
2. Principles of statistical graphics
3. Teaching
4. Making effective graphs
5. Communicating variation and uncertainty
6. Displaying fitted models
7. Giving a presentation
8. Dynamic graphics
9. Writing
10. Collaboration and the scientific community
11. Data processing and programming.
The course is set up using the principles of active learning, with class time devoted to discussions, collaborative work, practicing and evaluation of communication skills, and conversations with expert visitors.
The primary audience for this course is Statistics Ph.D. students.
STAT 6106 should also be ideal for Ph.D. students who do quantitative work in other departments and schools.
Here’s the syllabus for STAT 6106 this fall.
I’m so excited!
We’ve had this discussion before. For me, I have almost the opposite reaction. Despite being a happy classroom-based instructor for most of my career, I am happy to be totally online this fall (as I’ve been for most of the past decade). I don’t want to overstate it – the live in-person experience is highly valued and has its place. But the notion that all of the course (and 3-4 hours per week) need to be in person strikes me as odd and ill-placed. In fact, I believe that the in-person experience is enhanced if used more selectively than the traditional all in-person format. It does vary according to the students, however, and my feelings reflect older students, particularly working adults. But the traditional format where all learning takes place in the classroom never made much sense and I’m glad to see it disappearing. Further, all learning never took place in person anyway, so all I am really saying is that it make sense to rethink what should be done in person and what should not.
Dale:
In my class, most of the learning takes place at home, when students read the book and do the homework. My rule of thumb is that a class should take, on average, 10 hours a week of a student’s time—this is based on a load of 4 classes and 40 hours a week representing studying full time. So 3 hours in class is just 30% of learning time. I want the class time to be active, so I don’t do lectures and slides.
+1
I am loving the online mode instruction. I think it was an improvement for the class I teach.
I’m planning to do a data communication class in the winter aimed more at undergrads. I look forward to reading your syllabus. Mine was going to focus primarily on visualization in scientific communication.
I thought I was alone in not liking the Napolean-in-Russia graph
Huh?
Sorry, that was a bit cryptic, If you look at the syllabus for Course on Communicating Data and Statistics (last link), you’ll see the inclass assignment for the first week:
4. (5 min) Why I hate the Napoleon-in-Russia graph
Great idea for a blog post. Or… maybe it would just kind of ruin the surprise in the class? Or maybe the “hate” is conveyed in a way where he actually is saying how much he likes the graph but argues from a visualization rules sort of way that it’s terrible.
Andrew! Clarify!
:)
I hope it’s not ironic. I’ll feel alone again. People keep quoting Tufte to me about how it’s the greatest ever and I just don’t like it (yes. I can recognize the accomplishmentnof getting all those dimensions in a single graph) but it always struck as an example of cleverness over clear communication)
Google *statmodeling napoleon* and you’ll see two posts: “Against the myth of the heroic visualization” and “I wish Napoleon Bonaparte had never been born.”
I have that graph framed and hanging on the wall in my office (I got it years ago when I attended a lecture by Tufte). I like it partly because I’m Russian by heritage, though born in the USA. I find that very few of my co-workers ever notice the graphic or ask me about it. That’s a little surprising to me, because if I wasn’t already familiar with the graph, and if I had seen it displayed in someone else’s office, I definitely would have taken an interest in it.
Re Casual regression; is there any other kind?
There might be the laid-back regression or the multi-tasking regression, but those are likely to be the extremes of the distribution.
“Re Casual regression; is there any other kind?”
I can’t imagine that being log transformed feels “casual” to the data set! :0 Having your coordinates pumped through a ruthless and unfeeling equation and spat out the other side as new values sounds painful and unjust – brutal even!
Think of log transforms like taking off your rose-tinted glasses and seeing reality for what it is ;). Oops, that significant effect was due to two out of 500 data points. Redirect paper to file drawer or Frontiers. Or keep glasses on and submit to Cognition!
Do you have a link to the remote mic thing? Might try to buy one too if not too expensive.
Practical question: which mic/speaker combo did you get? This is the second person who’s mentioned one today, although that one was for teaching outside, a more likely teaching mode at UVA than at Columbia.
After a bit of browsing, I’ve found that the appropriate search term is, I think, “(portable) voice amplifier.”
Dl, David:
I ordered something off the web. I won’t mention the specific brand because I have no idea how well it will work in the wild, so I don’t want to give an implicit endorsement. Let me just say I did what anyone would do, which is to go to the site of a major online retailer, search for *remote microphone and speaker* or something like that, and then buy one of the reasonably-priced items that was also in stock. (Many of the top choices were out of stock, which I guess makes sense.) I can report back when I see how it works in class.
I think I have figured out online teaching. I recently taught a summer school course online and the response was overwhelmingly positive (https://vasishth.github.io/IntroBayesSMLP2021/). I feel that online teaching has huge advantages for students, and it took the pandemic for me to fully understand this point. But one has to structure the lectures very tightly—no improvisation, except in things that one is highly practised in. Stream-of-consciousness stuff (“here’s something I am working on right now and I will walk you through what I’m doing”) is worse than useless.
I’m of course a great fan of Andrew’s writings and insights, but I don’t like the way Andrew structures his lectures and presentations–actually, he hardly structures them at all! They are too stream-of-consciousness for my taste, and I think that people, especially beginners, would get a lot more out of his insights and knowledge if the classes were much more tightly structured. I still remember the early years of my statistical life, trying to understand what Andrew was saying, and failing. I think I now know how to explain his ideas in applied settings, but it took a lot of structured coursework and study.
I’m basing my comments on having heard Andrew’s lectures, and the syllabi listed above, which IMO focus on things that don’t bring a student forward in terms of actually being able to do stuff in real life. E.g., if you are teaching writing, it can’t be a single topic, it has to be the topic of a course. There’s a lot to learn there.
Also, Regression and Other Stories is not a book I would ever use for teaching an intro course; it’s probably more interesting for people who already know a lot of the stuff already. I like our own textbook much more, but I could obviously be very biased.
At least that’s my impression, could be wrong. I obviously don’t know what Andrew’s students say about the way he lectures.
Shravan:
When I’m invited to give a talk, I lecture. Anyone who’s curious as to what these lectures are like can find a bunch of them on Youtube. Some of my favorites are the talks I’ve given for the New York R conference. Usually the lectures go well, but not always. I don’t think the point of such lectures is to directly bring a student forward etc. as you put it; I think it’s more to get people thinking.
When I give a course (like the two courses discussed above), I don’t lecture. The courses are student-centered and highly structured with lots of activities for the students to learn by doing. A 13-week course and a public lecture are completely different things.
Regarding textbooks: there are many ways to teach, and I recommend that, when you’re teaching, you use a book that you’re comfortable with. It’s good to have different sorts of textbooks. One reason we wrote BDA, Regression and Other Stories, and our other books is because they’re unlike any other textbook out there. Different teachers have different styles and preferences so it’s good to have different sorts of books.
Yes, of course; I agree with everything you write. When I give a lecture, I like to tightly structure it so that the points I’m making are super clear.
Regarding teaching, it could be that students in Columbia are way more independent than in Potsdam. In Potsdam, I feel that I have to give a highly structure lecture, and then give them homework to apply what I taught them. It doesn’t work to just to do student-centered activities.
One thing I noticed was that students get very upset if they get something wrong; so what I did was I stipulated that if a student submits a complete and serious attempt to do the HW, they get 100% no matter whether the answers are right or wrong. Apart for a few edge cases where students didn’t submit the HW at all, but still claimed that they made a serious attempt and therefore deserved 100% (“I just couldn’t do it, but I did make a serious attempt”; I paraphrase), the students relax and try to just do the task to the best of their ability.