Flights of less than 350 miles should be a train ride. 🙃

The new Dungeon Crawler Carl comes out in 2 days

inside me are two wolves: - the wolf that thinks we need to show students challenges with NHST inference in huge (millions+) datasets and

You can’t just “do AI” if you don’t have good data and good documentation.

My grandpa: “I have found that the only way to goad busy numerical analysts into action is to claim that a subroutine is infallible” 😂 I guess the whole “the quickest way to get an answer is to post an incorrect on on the internet” was true even pre-internet 💀

Me: complaining that Canva docs don’t support in-line LaTeX @tonypelleriti.bsky.social: maybe they have a latex allergy

the only good thing* about having to use the delta method is getting to make @drob.bsky.social’s “(Taylor’s Version)” joke when writing it up *besides the fact that it “gets me” the “answer that I need”🙄

🤯 just had a wild experience. I needed to choose a hyperparameter value, 10 had good vibes so I used 10 for now. Later wrote a simulation that showed 10 is exactly when diminishing returns kick in (> 10 gives no benefit, < 10 shows poor performance). I think my brain subconsciously knew this 👻👻

WHY DONT WE HAVE RANKED CHOICE VOTING in CA Gubernatorial primaries?! (Or something even better, I’m not super educated on all the choices)

I am SO DELIGHTED that public libraries exist🥹 I went to return a book (operation bounce house) and pick up a new one (bobiverse #2: for we are many) and I was so HAPPY to see so many people there enjoying books, services, or just somewhere quiet to exist without buying anything 🙏 Truly incredible.

Me when I’m hoisted: surely this isn’t my own petard…🤨

I miss the “less is more” energy that generative AI work had in 2023/2024🥹 (quantization, specialized models, heuristic models with AI supplementing where it was actually useful) This is just a personal VIBE I’m getting, but now that genAI is more widespread it feels like (1/2)

data that’s clean, hallelujah. MCAR, hallelujah. bayesian priors, hallelujah. stakeholders on board, hallelujah. narrow CIs, hallelujah. ROI > 1, hallelujah.

Have you ever thought “this horror book doesn’t have enough statistics in it?” 🤔 Then have i got a book for you! Statistical, absurdist, horrific, and a little gay😚👌 this was a fun read. www.goodreads.com/en/book/show...

Ooop someone just accidentally reinvented meta-analysis again🤭

Nothing is certain except: - death - taxes - posterior intervals with overly narrow priors

One thing that’s stopped me from dreaming about being a Data Science Influencer™️ is that it seems like there’s so much pressure to put out such a VOLUME of content that you can’t make it all thoughtful/correct🤔 It’s just so effortful to put out content that’s helpful, original, + well thought out

the fact that @overleaf.com doesn't know Frequentist is a word is...interesting👀

Kinda peeved that out of all the cool work I did today the most impactful piece was a descriptive boxplot

Me when I’m writing for a respectable, very professional company so I can put my shitty jokes in 😒

YOURE NOT GOING TO BELIEVE THIS. I set out a mug + tea bag of what I THOUGHT was lemon balm, and @tonypelleriti.bsky.social boiled water and made me the tea. TURNS OUT I HAD ACCIDENTALLY GRABBED black tea. Instead of a calming bedtime libation, I drank 50mg of caffeine @ 11pm 🫠

Me drinking my bedtime Lemon Balm Tea in case it will magically cure all my stress and anxiety 🙂🍵

This is another great example of a rule that resonates with me: When someone asks you to send more time consuming/interacting with material than was spent creating it, it is 🫩SOUL CRUSHING🫩

I used to think it was so quaint in recency novels when ladies would “catch up on their correspondence” So tell me why I just spent 30 minutes responding to various unread messages that have accumulated the last few days😂

I feel bad for medical doctors. I would hate watching so many popular shows about statistics and seeing them get everything wrong 😵‍💫

How privileged are we to understand math as a language??? It’s so much more difficult to explain complicated topics to without this shared language 🥴🫠

Bayesian SaaS companies and OnlyFans creators are very similar if you think about it🤔 Both make money by showing off their expertly crafted posteriors.

Not this historical fiction teaching me how to git rebase 💀

Am I promoting badly or do LLMs seem quite bad at statistical interpretation and causal inference 👀 (I’m not asking it to run models, but just thinking through design/inference/implications)

Just had such a nice reminder today about how far I’ve come since pivoting from psychology to statistics 🥹 I wish I could go back and tell 2016 Chelsea that one day she will not only have deep stats expertise but other people will (sometimes lol) listen to her and respect her opinions on stats😱

Start calling literature reviews “the lore”

Using coding agents sometimes feels like herding feral toddlers

You’re telling me a cat displayed the contents of this file to console??? 🐱

workin on some stuff

Therapy for anxiety has accidentally made anti-consumption much easier! When I feel panicky, my therapist and I practiced tolerating the feeling just for a little while, then checking in and seeing if I could tolerate it longer.

The rise of AI slop across all domains, including its disturbing (yet unsurprising) penetration into academia, comes with a nice twist for us who reject it: Having actual domain expertise, a deep understanding of a field, a firm grasp on a topic... These now seem like magic 🪄, like a superpower🦸.

My favorite method of scicomm😈 jokes that make people so curious they have to learn a little to fully appreciate the meme.

ggplot defaults are bouba, matplotlib defaults are kiki

A math class, but it’s just me teaching something super in depth with the SOLE PURPOSE of students understanding one very niche meme I made by the end of class

If you’re having eigenproblems I feel bad for you son, I’ve got 99 eigenvalues…but the residuals for my unconverged Ritz values are no long steadily decreasing so i need to restart my Lanczos algorithm with a new starting vector 😔

The problem with math education is that eeeeeveryone and their brother wants to make a video on how to calculate something sexy like a determinant, but AS SOON AS YOU want to learn about krylov subspaces there’s like one guy who made one video about it 10 years ago take it or leave it 🤨

Me when Claude calls another one of my points “sharp”

Me when someone tries to distract me from generating shareholder value

How I go about my workday knowing I’m going reward myself for being a good little statistician with a crispy fountain Diet Coke today 🥤

THE WORDLE CREATOR MADE A CRYPTIC CROSSWORD GAME I AM SO HAPPY www.parseword.com

The CLT finally showing up when n = 30*

It’s surprising to me how often the solution to my problem is “randomize/analyze within segments rather than pure randomization/analysis”🤔

The most fascinating applications of statistics are the ones where you have to creatively work around pragmatic constraints. E.g. quasi experimental methods.

Mon mari m’a acheté une livre de Python en français pour mon anniversaire. Bien que je connaisse bien Python…le français…BEAUCOUP mois…😂 alors je vais apprendre rapidement.