How it feels the few months after your defense

THIS is what linear algebra does to your brain, and they teach this stuff in SCHOOLS?!?😱

Im a data scientist, I love data scientists, but we should not be allowed to name things anymore 🙅‍♀️

If you’re going to use LLMs to help you build and app or code: 1. you need RIGOROUS, rigorous unit testing, and/or 2. you need to already mostly know how to build the thing yourself so you can check and modify I fear #2 is going to become less common, so let’s be super adamant about #1 😬

I fear I’ve lost touch with the common man when it comes to statistics 😔 I need to get back in touch with my roots and make some like FROM THE GROUND UP stats materials, I’ve gotten to stuck in the weeds😂

TW: food tracking I track my food intake (macros, fiber) sometimes, and the app often flags random foods (chia seeds today) and flags them as associated with higher or lower calorie days. Surely this is FRAUGHT with multiple/sequential testing issues?? 👀 wonder if they account for that

Hey has anyone else noticed that augmented synthetic control packages (at least in R; geolift, augsynth, tidysynth…etc) have very little documentation compared to what I would expect given their usefulness???? (Ask my why I know🫩)

Kindle books often show the average time to read an ebook when you first open it. If I were in charge I’d definitely show the median time not the mean time 🤔 and I feel like that’s a good example to use in the classroom.

It is so incredibly disgusting, horrific and unsurprising that ICE has killed another person. I don’t even know how to publicly talk about this stuff anymore. The descent into fascism is swift and is accompanied by an incongruous "normality" in your day-to-day (for the privileged). I’m sickened

How am I supposed to order a “pain au chocolate” as an American without sounding like I’m trying to be so fancy using a French accent🥐😩

Me when anyone asks for me to justify my analysis methods

Two random (🥁) inference thoughts: 1. you cannot just hand wave binary (go/no-go) decisions and pretend like error rates don't apply to you if THE USERS ARE GOING TO MAKE A BINARY DECISION WITH YOUR ANALYSIS

Either Causal Inference in R www.r-causal.org or The Effect! www.theeffectbook.net/index.html

After 4 months, have I had an impact at Canva? On one hand, I’ve made tools and done work to help people make better marketing decisions. On the other hand, no one has made a custom slack emoji of me. So who can really tell.

I’m proud to say I used SOCAHTOA yesterday 😌 @tonypelleriti.bsky.social and I were choosing crown moulding for our bookshelves and calculated the adjacent side length (whether it would fit btw bookcase and ceiling) based on the hypotenuse (mounding height) and angle😌

MMMs and other marketing models often assume strictly non-negative ROI on ad spend but… perhaps they shouldn’t… 😂💀

If the value of an event is big enough in magnitude, even at very low probabilities it’ll have a HUGE impact on your expected value (This is a post about anxiety)

Fascinating and perhaps (?) an interesting example where mixture models like a zero inflated poisson or negative binomial distribution could be useful 👀

Selecting hyper parameters without grid search, just to anneal something

Mathematicians really add a single diagonal to a matrix and completely change what it’s called 🙄

The horrors persist, but so does my writing. I reviewed all of my writing this year over on Patreon, www.patreon.com/posts/146964..., including a few pieces that until now have been Patreon exclusives.

Tfw an operation takes a sparse data structure and makes it non-sparse

whenever (X^TX)^-1X^Ty comes up while teaching, I always mention a fun fact that I've met (virtually or otherwise) not one but TWO people with this tattooed on them.

You know what they say about sparse matrices? They’re mostly a whole lot of nothing.

You know what they say about large eigenvalues…😏 …they explain the most variance.

This year I tracked how much money I'm (theoretically) saving by using my local library (I read a lot, especially when stressed) 📖 I read 100+ books this year, and 71 were from the library (physical or digital) 🤯 If I had bought all those books form @bookshop.org it would have cost...$1,506.87

I’m gonna say it…I don’t care if your causal inference methods need to be more complex in order to protect consumers from being price gouged based on their individual characteristics in the future

Im rich in life because people like when I make jokes like this 🥹🥰

I suppose my clothes DO locally resemble real n-dimensional Euclidean spaces 🤔

computer going to take my job? no. i’ll pour water on it

RBF for RBF: using support vector machines withto classify resting bitch face

📢 TEACH DATA SCIENTISTS TIME/SPACE COMPLEXITY FOR COMMON ALGOS AT INFERENCE/TRAINING!! EVEN IF THEYRE NOT DOING MLOPS-Y STUFF, IF YOURE WORKING AT A COMPANY WITH HUGE AMOUNTS OF DATA, IT WILL BE HELPFUL TO KEEP IN MIND WHEN DESIGNING ANALYSES. #pikasshoutstats

I don’t get it, why is almost all “code” stock images/ b-roll in HTML?!?!?

You smug little…

You know that thing Coco Chanel said about accessories? “Before you leave the house, look in the mirror and take one thing off” We should all do that but with about 50% of the text we put in slide decks. TAKE. IT. OUT.

There’s a guy on tiktok who’s trying to replace ChatGPT by first drinking a ton of water and then answering people’s questions poorly 💀

Yes! I think we do need to be more clear about the benefits you get from being in the right hand column vs the benefits of being in the bottom row !!!

I just fell in love with a sci fi book and found out it has 7 books in the series (so far) so unfortunately my social life has wrapped up for the year, if you need me I’ll be reading ✌️

👀 Green bean casserole spotted, ✅ green bean casserole acquired 🐶

I think I’m about to hyper fixate on making the best focaccia ever. I tried a brined focaccia and it was 🤤

Braiding Sweetgrass (paraphrasing): no one becomes a scientist because of the wonder of data or p-values Me: ☝️ not NO one…👀

Oh no I have a query that will take ~7 minutes to run, I can’t possibly do anything productive during that small amount time, I better go scroll on social media…!

what are those ✨nostalgic✨ textbooks for you? not the most useful textbooks, or the ones you liked the most, but the ones that really take you back... for me, Strang's Linear Algebra and Casella + Berger's Statistical Inference come to mind

That’s why we have strongly informative priors amirite? /s

Sometimes people just want numbers that tell them to do what they already want to do 🤷‍♀️

📈📉

IF YOURE NOT GOING TO LOOK AT THE UNCERTAINTY OF YOUR ESTIMATES WHY BOTHER DOING THE ANALYSIS AT ALL #pikasShoutStats

She doesn’t even know about the assumption of homoskedasticity 🥹

The people yearn for adorable math slides

Me: *unsuccessfully tries zero-shot learning* Me to my LLM: