@carlislerainey
political scientist at FSU; experimental design, inference (frequentist and Bayesian), metascience Web: https://www.carlislerainey.com Google Scholar: https://scholar.google.com/citations?user=otXLf3
I think three things are almost certain to happen when one writes in a new area:
1. The paper is more innovative, which makes it harder to sell
2. The paper is worse, because one doesn't know the conventions
3. The paper takes longer, because everything is new
4. Others struggle to find the work
Put together a quick starter pack for political methodologists (broadly defined) who have made their way over to BlueSky. If I've missed folks, please let me know!
go.bsky.app/JBavA7x
Would you like more than one? I donโt mind.
Don't forget to get your #SPSA2026 proposals submitted today!
The deadline is today.
It's January 14-17, 2026, in New Orleans. And it's always a great time.
spsa.net/annual-meeti...
This paper (and its working versions before it) have probably done more to reorient my thinking about design than anything else I've read. Just really critical work for political science to think about in any type of hypothesis testing environment. Well done to this team.
I'm chairing methods this year. I'm looking forward to putting together some great panels. Submit your papers!
Deadline is Sept. 15!
I'm chairing methods this year. I'm looking forward to putting together some great panels. Submit your papers!
Deadline is Sept. 15!
Steal away!
Here's the idea.
Someone pops into your office and says "Hey, can you tell me about X?" You say sure--you give a 10 minute pitch. You describe the motivation, basic idea, and suggested resources.
I want to write these short pitches down. Here's a first one on power.
๐ ๐๐ฏ๐ฆ-๐๐ข๐จ๐ฆ ๐๐ณ๐ช๐ฎ๐ฆ๐ณ ๐ฐ๐ฏ Statistical Power
www.carlislerainey.com/blog/2025-08...
Excellent reference for those, like me, who can always benefit from a refresher on statistical power
New Post: "For Your Syllabus: Statistical Power"
Add content on statistical power to your social science courses.
Not just to methods courses.
For substantive courses, Bloom's MDE (i.e., 80% power to detect 2.5*SE) is easy to teach and really helpful!
www.carlislerainey.com/blog/2025-08...
New Post: "For Your Syllabus: Statistical Power"
Add content on statistical power to your social science courses.
Not just to methods courses.
For substantive courses, Bloom's MDE (i.e., 80% power to detect 2.5*SE) is easy to teach and really helpful!
www.carlislerainey.com/blog/2025-08...
โผ๏ธCool new paperโผ๏ธ
Finds that journal data policies in psychology boost sharing statements to ~100%, but only about half of datasets are complete, understandable, reusable.
Open: open.lnu.se/index.php/me...
An old boss once told me โeach sentence should carry one idea and one idea only.โ Good, Strunk and White-esque advice.
Love this. You clearly describe how much writing is a *struggle* relative to the often-used description โwrite it up.โ
Whatโs the single best piece of feedback youโve ever received on your writingโsomething that stuck with you and shaped how you write future papers?
Observations:
- A surprising amount of good ideas emerge from the small groups. Thatโs where the magic happens.
- Students who tend to avoid talking to the whole class seem to share more freely after building confidence in the small groups.
Iโd say pose a question to the class, ask students to brainstorm as individuals for 1 minute, discuss in groups of 3 for 3-5 minutes, and then discuss as a whole class.
Related to this new paper, in case you havenโt seen it yet.
GitHub: github.com/ArthurSpirli...
Gated JOP: www.journals.uchicago.edu/doi/10.1086/...
Assign students to ask class-related questions and grade them on the quality/depth of those questions. Provide the feedback and use the questions to help shape what you do in the classroom. Hard to scale above 35-40 students, but has been super helpful to me.
But to your point, we should require โcareful research designโ of both descriptive and causal papers. Certainly designs for causal claims have tricky parts, but designs for descriptive claims are not trivial.
But different designs are required to defend causal and descriptive claims (and least to defend them well)โdescriptive and causal papers must (usually) use different methods.
(And by โthe empirical design,โ I donโt think they mean the presence/absence of inferential statistics.)
I think we agree then!
Whatโs your preferred label for the claim โaffective polarization has increased in the US over the last 20 yearsโ?
Iโd call that a โdescriptiveโ claim. What do you recommend instead?
1. We have a tendency to make things easier for students these days, but the more you demand, the more you will get back from the students (on average).
2. There is no teaching style that is best. Find what works for you and improve it over time.
1- Use real world examples.
2- Unpopular opinion: done in moderation, edutainment is not a bad approach. It breaks monotony and maintains engagement.
1. Increasing the number of assigned readings typically has little if any relationship to the extent of reading that will happen, so focus on interesting assignments if you want to add to the course expectations.
2. Donโt be afraid to say โI donโt know but Iโll look into it.โ
Every single time, ask yourself what is the point of THIS talk to THIS audience, and then do something with the answer
