Associate Professor,
@Harvard
. Political scientist researching statistical methodology and US politics. he/his also at
@matt_blackwell
@mastodon
.social
Reminder that I have a textbook on an introduction to mathematical statistics and regression, designed for first year PhD students in poli sci, but maybe useful to others. Let me know if you use it and have feedback!
After a long and unsuccessful search for a statistical inference and regression textbook at the right level for our PhD students, I finally decided to write up my lecture notes as a book.
The book is publicly available if you are interested:
Interested in causal inference? I updated my course last year and have posted the course materials online. If you're teaching a causal inference course, please feel free to steal my lecture notes!
BREAKING: Notre Dame President Fr. John Jenkins, who was at the WH SCOTUS announcement on Saturday and was criticized for not wearing a mask and shaking hands, has tested positive for COVID-19.
This was just sent out to the campus.
Unclear if he had it during the WH event.
My best tip on how to give better quantitative presentations is to (a) use more plots and (b) build up your plots on multiple overlays, as in:
- Just x-axis (explain it)
- Add y-axis (explain it)
- Add 1 data point (explain it)
- Plot the rest of the data (explain it)
Honestly, I’d much prefer a SCOTUS Justice that has basic statistical literacy than one whose brilliance in con law allows them to write beautiful dissents or the mind-blowing 14th prong of some new test.
Last semester I taught an undergrad quantitative methods class and decided to pre-tape most of my lectures. They're not the highest production value, but I thought maybe someone might find them useful so I made a YouTube playlist to share them:
I cannot believe a serious moral philosophy has been built up around the idea that humans can predict the consequences of our current actions on the far future with any degree of accuracy. Insanity.
For those teaching quantitative methods classes in the social sciences: I put together a {learnr} package called {qsslearnr} with interactive tutorials that correspond to Kosuke Imai's QSS textbook.
One of the more subtle manifestations of my imposter syndrome: feeling slightly relieved as a reviewer to find out the other reviewers had the same reaction to the paper.
One R function that I forget about and maybe you do too: reformulate() can take a character vector and turn it into the right-hand side of a model formula (and you can add the outcome too). eg:
reformulate(c("var1", "var2"), response = "outcome")
OLS is out here every damn day, using every muscle to pack those residuals as tightly as it can. And then you have the audacity to complain about the result? Oh, is it too “variance-weighted” for you? Boohoo. Go home and cry to your sample mean.
@PhuzzieSlippers
Yes! It completely reoriented my thinking on slavery when I realized, to whites at the time, slavery was like homeownership today: the path to upward mobility. Absolutely zero regard for the humanity of the enslaved.
If I’m reading correctly, Econ has found the One True social welfare function and now it’s just pew pew pew optimal policy all day long. Love it, congrats everyone!!!
Last Fall I completed another iteration of Gov 50 (Data Science for the Social Sciences), and I've been really happy with the course. We doubled in size to 250 students and got great reviews from students. What's worked for us?
If I could give a single piece of advice to new PhD students (in poli sci at least): work as hard as you can to write a complete, submittable solo research article by the time you start your 3rd year.
Even if it is never published, the experience you get is insanely helpful
The entirety of one of the best Econometrica papers. Certainly the best to contain the phrase "ceiling, celery, ceremony, cease, cedar, celestial, celibacy"
In case it’s helpful to folks, I have a set of lecture notes available for a grad-level causal inference course here: (with great section notes from the great Sooahn Shin)
Feel free to use any of the material in your courses. Happy to share sources!
Some random practical PhD advice in no particular order:
- almost everyone should start their dissertation by writing a paper, even if you plan to write a book dissertation. It’s less pressure and you can always expand it if you find the page count growing and growing.
This Von Neumann quote should probably be on every syllabus of every remotely mathy class: “in mathematics you don’t understand things, you just get used to them”
I was curious about the data and so I played around with it and found another teen trait that has changed dramatically over the last 10 years: sleep.
Since 2014 we see a fairly steady decrease in the proportion of 12th graders saying they sleep 7hrs ever or nearly every night
There was no sign of a teen mental illness epidemic until around 2012. Then liberal girls' rates started increasing. Then everyone else. Why?
@glukianoff
nailed it: Reverse CBT, as I explain here:
One practice that is useful to begin in grad school is empathetic criticism, in which you try to give feedback from the target’s own perspective. Instead of saying X is dumb, try to understand the trade-offs that led the author to choose X and tailor your comments accordingly.
I'm in awe: the 90% effect arm of the Oxford UK study wasn't in the original protocol because that dosing was done in error and they just rolled with it.
When working on an R&R, I often dream of being able to have an anonymous text/email exchange with the reviewers to help me understand what the hell they are talking about.
🚨 New paper 🚨
In an experiment when should you measure a key moderator? Before treatment risks priming bias, but after treatment risks posttreatment bias. In a new paper, we develop methods to bound each type of bias in any experiment.
🔗:
This might be my dumb statistics brain speaking, but wouldn’t it be a good idea to, say, randomly sample 1000 people in the hardest hit states to get a sense of the infection prevalence? Do we not do this just because of testing capacity issues?
A “wirecutter for statistics” site:
Best Regression Estimator for Most People (OLS, upgrade pick: LASSO)
Best Regression Variance Estimator for most People (Robust SEs, budget pick: homoskedastic SEs)
This Haidt article on what has caused polarization in the US spends 2 paragraphs on rightward drift of Republican party and 13 on the identity politics of universities
I love creating simple animations for teaching with my beamer (pdf) slides and I always found it to be a huge pain in Rmarkdown, which I otherwise love. So I finally sat down to fix it and it moved me to blog for the first time in ~8 years.
Gonna say, probably not great that we live in a gerontocracy imposing policies that are very unpopular with young people with seemingly no democratic way to reverse them.
Excited to announce my new venture, Prior, a platform for all of the ideas you already believe written by the writers you already love.
For the low low cost of $500, we will make sure you feel completely justified in your beliefs.
At Harvard orientation we did an icebreaker where we had to declare our favorite sandwich estimator. I first thought “CRSE with HC2” but I didn’t want to seem too nerdy so I said “White SEs” & a person said that has terrible finite sample performance and everyone started snapping
I've been teaching a revamped version of our intro to methods/data science class for undergraduates, Gov 50, this semester and it's been a lot of fun. More live coding in lecture, more simulation, fewer formulas.
Details here:
Agreed: don't reinvent the wheel! Here are materials from my recent classes:
Grad Probability/Inference/Linear Models:
Undergrad Data Science:
Grad Causal Inference:
Feel free to take as much as you want!
Folks, this is a really hard time to be prepping classes. I have materials for Intro American Politics, Media & Politics (needs updating), Grad scope & methods, Grad Surveys & Experiments, undergrad data analysis. If I can help you prep, let me know.
Also out today: my paper with Adam Glynn on causal inference in TSCS in the APSR. We try hard to clarify the quantities of interest and the possible biases that arise in this setting.
Extremely happy and humbled to report that my book with
@maya_sen
&
@aviditacharya
Deep Roots has been awarded APSA's 2019 Riker Award for best book in Political Economy in the last 3 years.
Slides are often an imperfect teaching tool and my students will definitely tell you that I had my fair share of equations, but I really enjoyed the challenge of creating intuitive visuals for key concepts in our probability+linear regression course this semester.
The authors said by email that they used a built-in Stata function and aren't sure themselves how the software used the input weights. I suspect they misapplied that function (too complicated to tweet why) but I don't know Stata well enough to be sure; it seems neither do they.
I think academics have taken their eyes off the ball in the debates over academic freedom.
So much ink spilled lamenting students shouting down speakers while the right is trying to dismantle the tenure system and shutter academic programs based on ideological grounds.
Even though the list hasn't been updated to reflect it, I signed this letter saying that I won't retaliate against striking student workers at Harvard.
As usual, 3Blue1Brown does a fantastic job stepping through the basics of (parts of) the GPT algorithm. One of the best introductions to the idea of embeddings that I’ve seen
In principle, I don’t believe anyone should own or run Twitter. It wants to be a public good at a protocol level, not a company. Solving for the problem of it being a company however, Elon is the singular solution I trust. I trust his mission to extend the light of consciousness.