Political Scientist. Professor at NYU’s Center for Global Affairs. Experiments, data analysis, guitar, drums, fan of comedy. Two boys and exhausted all the time. More at www.johnvkane.com
Genuinely *beyond thrilled* to announce that I have just signed a contract with University of California Press to publish my book, "Truth Addict: How to Start Thinking Like a Social Scientist in a Data-Driven World."
It's a true passion project & I'm so excited to be writing it! Wish me luck! 😁
Thanks so much, Zander! And yes, the bars are really unnecessary (though maybe for non-quant folks it makes reading the graph easier? 🤔).
In any case, -coefplot- can very easily just plot markers with CIs rather than bars. I’ve even done this myself in a few papers…
Also FWIW, we presented and framed these shifts as being "huge", AND a manipulation check found that respondents did indeed view these to be pretty large shifts.
But still, it's an open question whether, say, a group's change of 40% and/or being majority Republican, might find an effect.
🚨New working paper w/ some terrific Stony Brook grad students!
Using a pre-registered experiment, we test whether shifts in groups' voting behavior can change partisans' sentiments toward members of those groups (Black, LGT & Muslims).
Answer? No.
polisky
preprints.apsanet.org/engage/apsa/...
🚨New working paper with Sam Perry!
Do religious beliefs affect beliefs about climate change? If so, which beliefs, and why?
We argue & find evidence that belief in divine (vs. human) control over Earth's fate lowers concern about climate change. Preprint here: doi.org/10.33774/aps...
polisky
Many thanks to the great Erin Rossiter for inviting me to present to her wonderful group of PhD students at @notredame.bsky.social! It's truly a pleasure to talk about null results in experiments!
If you'd like me to present to your students, this or next semester, email/DM me! 😁
Thanks Stata & @asjadnaqvi.bsky.social 's schemes for helping me make some nice graphs for my son’s 2nd grade science project!
The artwork is all my wife. I *may* have had a hand in the punny title. My son created the data set (and…ate the candy).
p.s., null results have never been so delicious.
Right?! He’s given me so much amazing feedback and has been so unbelievably complimentary toward stuff I’ve done…but then right when I feel a little bit of confidence, I realize that his 19th publication of the month just came out in the PNAS or Nature or whatever
He’s just a legend ❤️
Had the honor of reviewing Jamie Druckman’s book, “Experimental Thinking” for POQ.
If you do experiments, get this book. Aside from the fact that Jamie is an incredible (and kind!) scholar, the book has more wisdom-per-page than anything I’ve ever read.
academic.oup.com/poq/advance-...
🚨Stata folks: Do you or your students ever want to make "stacked" bar graphs in Stata?
In Pt.3 of my guides on bar graphs, I cover how to make a variety of stacked bar graphs & techniques for making them prettier. Includes all code 👍 Enjoy! 😎 #StataGallery polisky
medium.com/the-stata-ga...
The goal was to identify all the things that could explain a null/non-significant finding in a survey experiment *other than* a theory/hypothesis being incorrect. If you can rule out these 7 "alternative explanations", your null results become much more informative. 👍
Do you or your students ever need to make bar graphs in
Stata?
If so, I just published a "Pt.2" of my guides on advanced bar graphs, this time covering visualizing cross-tabs, -catplot- & adding a "Total" category. Includes all code. Hope it's helpful for you! polisky
medium.com/the-stata-ga...
Ever need to make bar graphs of means in
Stata? Ever wish you could add confidence intervals?
If so, I made a new guide in the #StataGallery on
@Medium on how to do it using -cibar- and/or -coefplot-. Lots of examples, tricks & of course all of the code. 😎
medium.com/the-stata-ga...
Across a variety attitudinal & behavioral outcomes, the STW treatment increased Republican openness: Republicans were more willing to get the COVID vaccine/booster, more likely to encourage family to get it, and more willing to request
CDC facts & info about where to get it.
Do you use @Stata for regression analysis?
If so, I wrote a guide for #StataGallery on @Medium that covers not just how to make nice marginsplots for visualizing results, but also how to *combine multiple marginsplots* into one graph. Hope it helps! 😎 polisky
medium.com/the-stata-ga...
Looking at US pres. election data for an R&R and am fascinated by the recent decline in margins of victory. Unsurprising in our era of polarization, but given the historical trend, suggests either a very close 2024 election and/or that another big political shake-up is coming. 🧐
Just wrapped up another wonderful semester of "Advanced Data Analysis for Global Affairs" at NYU's Center for Global Affairs. Such a fun course to teach, and 2 terrific students surprised me by wearing Stata t-shirts to our last class! 😂
This strikes me as a crucial result in the @nytimes poll. An *unnamed* Democrat wins huge in battleground states: better than Biden did in 2020 or Harris now. What does it suggest? The battleground states have not shifted toward Trump—they want a Dem but can’t commit to Biden (or Harris). Yet.
🚨Very happy that my co-authored paper with Ian Anson is out in the new issue of Political Behavior! The key finding is that, among citizens and partisan media, there is a strong tendency to care more about government deficits/debt when the OTHER party controls the presidency.
rdcu.be/dp3rD
Quick @Stata graphing tip: use the "vioplot" pkg to create violin plots. -ssc install vioplot, replace- These basically combine a histogram and a box-and-whisker plot into one graph. Esp. nice when showing a continuous variable across different categories. Check it out 😎
Quick @Stata tip I really wish I'd known sooner 🤦♂️: Ever feel like the size options for labels or markers aren't exactly the sizes you want? Just use "mlabsize(*#)" or "msize(*#)"! E.g, make a label 72% of the default with "mlabsize(*.72)" and a marker 3.12 times bigger w/ "msize(*3.12)"
Hey Stata folks! Want to visually present your regression results? I made a guide. It covers key aesthetics, combining results from multiple models, working w/ logit models, & more. Oh, and I provide all the code. Comments/feedback welcome! drive.google.com/drive/folder...
Lastly, I explore panel data to examine which voters might be more likely to vote based upon incumbency status of candidates (either 1-term or 2-term incumbency). Key takeaway: Less politically interested, less educated, less partisan citizens more likely to use this heuristic.
First, presidential incumbents (i.e., served 1 term) have an *inherent* advantage. The experiment isolated incumbency from every other potential confounding factor. Regardless of whether it's a D or R incumbent, the effect is positive: ~5.6 percentage points.
This experimental result lines up nicely with election data. Since 1952, candidates who have served one term win the popular vote ~78% of the time and average ~54% of the two-party vote.
What about candidates whose party has served two terms (think Gore in 2000, McCain in 2008, HRC in 2016)? Are they inherently disadvantaged? For example, are voters just ready for a change? Answer: No. They perform worse than 1-term incumbents, but NOT worse than the control.
