John V. Kane

I mean, his name is in my handle. I almost feel a moral responsibility 😆

Noem quoting Orwell is peak Orwellianism

Finally had some inspiration to play with the 'colorpalettes' package in @statacorp.bsky.social.

Tried out one of the Wes Anderson palettes (originally made for R, I believe) and now I feel like I've been missing out for the past several years... 😕

🚨 Interested in survey experiments? I'll be giving a virtual talk on (false) null results in experiments--and how to protect against them--at SWERP on 3/13 (3/12 8pm EST). Hope to see you there, and thanks to @sysilviakim.bsky.social for inviting me!

Link to join: swerp.netlify.app

Just finished this terrific episode. Well done! 👍

p.s., as a lifelong guitarist, your gripe with “Jump” was so perfectly spot on. But does Patrick know yet how many times you inserted that sample?? Has to be a new Quantitude record 😂

Ha! Plenty of articles in the pipeline (and a book, at long last!), but not nearly as much music as I’d prefer. Maybe that can be my New Year’s Resolution… 🤔

Thanks again, Jason! 🙏

Thank you, Jason! Really means a lot coming from an outstanding scholar like yourself! 🙏

You've played a big part in that trend line, so thank you for being such an amazing co-author, Ian!!!

🥳 Well, the small milestone has arrived. It's nowhere near many others, but it would've been impossible for me to imagine ever happening when I first started the PhD

With the kids having school canceled today & tomorrow after a 9-day vacation, this is a nice gift. Thx citers!!!

4 years of observational data plus three survey experiments. Lots of surprising findings here, including that priming norms (esp. seeing members of the OTHER party be unconditionally biased) can lower one's own probability of exhibiting UPB. 👍 Paper here: preprints.apsanet.org/engage/apsa/...

🚨Fully revamped paper with @iganson.bsky.social: We study levels of, and partisan differences in, "unconditional political bias" (UPB). This is a partisan's own admission that *nothing* a politician could do would lower their support for that politician.

I had the same thought. Hard to imagine how a 60 min job talk with Q&A will go if the candidate played no real part in the paper’s production. Not just for technical questions but, perhaps especially, Qs about all the stuff a candidate would have typically thought hard about while writing the paper

Had a 𝐰𝐨𝐧𝐝𝐞𝐫𝐟𝐮𝐥 time presenting on survey experiments & null results in @semrasevi.bsky.social's class at U of Toronto! Such great student questions!

🚨If you would like me to present in your course, please feel free to reach out--I truly love talking w/ others about this topic!😁

Thanks so much! This isn’t from a paper or anything—just an analysis I was interested in checking out. But if you think I could help in any way, please feel free to email or DM me—I’m happy to chat! 😊

Thanks! The model controls for age, but I didn’t separately look at how party interacts with age. I’ll try to give it a look, though, thanks!

Are Americans polarized in their attitudes toward higher education? 🤔

Using some @electionstudies.bsky.social data, the answer seems to be: Yes, but not dramatically so, and there's a tendency for more experience with higher ed. to be associated w/ more positive ratings. 👍

Exactly. One thing I keep thinking is that (maybe?) lower approval in the public serves to lower the costs of defection within the GOP. Get below 35% and I’d expect more Liz Cheneys, Mitt Romneys, and MTGs. They’ve been there all along, but didn’t want to alienate the base.

I will find it hard to pay much attention to his approval rating until it dips below this 35% mark. To me, that would signal a true shift.

But it still might be unlikely. For reference, Richard Nixon—Nixon!—left office at 24%. That’s probably the real “floor,” not 0%.

Beware the (Double) Axis of Evil: How Graphs with Two y-axes Can Deceive You Two y-axes in one graph? Here's why you should exercise caution.

Link to article: medium.com/the-stata-ga...

💡For those teaching stats, data literacy, methods--I just wrote a short Medium piece for @asjadnaqvi.bsky.social's
amazing Stata Gallery. The piece covers how and why graphs with two y-axes can be so deceiving. Includes an applied example (with code). Hope it's useful! 😁 Link 👇

Here's a screenshot of the code. Admittedly it's a bit more complicated than the simple "xline()" approach, but until
@statacorp.bsky.social adds something like a "top" option (like how "citop" exists for CIs in -coefplot-), it might (?) be the best way to do it...

Ever need to add reference lines in
@statacorp.bsky.social graphs? If so, you've probably noticed that the lines go *behind* bars/bins. 🤷🏻‍♂️

Today I discovered a hack to fix this. Instead of using e.g., "xline()", use "scatteri". Ref lines will go *on top* of bars, not behind. Code in thread 👇

Key point: when the true effect is positive, same-sign relations b/w Z & X, and Z & Y make effects too positive. Opposite-signed relations make effects less positive.

But when the effect is negative, this flips: same-sign→less negative, opposite-sign→too negative. 👍

Just as in classic textbook examples of OVB, here exercise (X) is getting too much credit--credit that should (at least in part) be given to low caloric intake (Z).

An example: More exercise (X) should ⬇️ weight (Y).

Suppose caloric intake (Z) is negatively correlated with X, and positively correlated with Y, but we don't control for Z.

The estimated effect of X on Y will be *too negative*: high exercise people are ALSO low-calorie people (and vice versa).

When the true effect of X on Y is 𝐧𝐞𝐠𝐚𝐭𝐢𝐯𝐞, and Z is positively related to both X and Y, omitting Z will make the slope *more positive* than it should be.

When Z has opposite relations with X and Y (➖X,➕Y), the bias makes the effect *more negative*. 😯

We all know about omitted variable bias: when X⬆️Y, Z is ➕ correlated w/ X & Y, & we omit Z, it biases the effect of X on Y upward (too positive).🥱

But what about when the true effect is 𝐧𝐞𝐠𝐚𝐭𝐢𝐯𝐞? Is OVB just the mirror opposite, biasing the effect to be too negative? No.👇

Looks amazing! Congrats, Nathan!!! 🥳

Hoping to coin a new term: parAInoia

Definition: The increasing feeling of dread (felt among teachers/professors while grading a paper) that a student's work was written, at least in part, by AI.

Example: "It took forever to get my grading done because I was feeling intense parAInoia."