Let 'Em Cook? Kicking off a series on rookie minutes and rookie growth

Trying to get back into writing. In my latest, I look at NBA rookie minutes and the Wizards and cacti. Like and subscribe!🌵 open.substack.com/pub/wizardsp...

If you're on the data science job hunt and feeling discouraged just know that there are terrible clustering algos out there, in production, and you can do much better. Like, look at these 'similar shoes' from DSW. If you're reading this, you can get better results. I believe in you!

Still hurts to see the Wizards at 6 after seeing the Wizards at 18-64.

Say what you will about Poole as a player, but he went from laughing stock to really winning fans over in DC this past season

GitHub - jacobpstein/psm_did_sim: Repo for PSM + DID bias from inclusion of treatment affected covariates in a matching model Repo for PSM + DID bias from inclusion of treatment affected covariates in a matching model - jacobpstein/psm_did_sim

In my old day job, I pushed for more simulations to inform causal design, check methods, and help us learn. I posted some code and did a little write up on LinkedIn with a colleague about a case of propensity score matching that crossed our desks. github.com/jacobpstein/...

Sometimes you accidentally write a recursive loop and that's when the fun really starts.

If only someone would write a whole book about this!!!

nickwan_datasci - transparency matters Watch nickwan_datasci's clip titled

What a difference a default makes: on viz buzz last week I used theme_minimal, which defaults to a clear background. My viz was 7% similar to target viz due to transparency @libbyheeren.bsky.social & @nickwan.bsky.social checked again with a white background and well… m.twitch.tv/nickwan_data...

Very cool data viz showing game flow in 3D for the NBA finals vsueiro.com/hoop-hills/

Trolli also had the sour Harden gummies back in the day

Thanks! So I think there is a serious selection problem in the data since Project Gutenberg doesn’t have access to a lot of modern literary copyrights. That said, I wouldn’t be surprised if the 19th century was the peak (at least adjusting for overall population)

I barely had any time for #TidyTuesday this week and want to revisit these Gutenberg data sets with some LLM tools at some point. I looked at life spans but kept it to the period since the modern novel was born. This could be a good interactive if I were doing a quarto presentation

Congrats on this! It feels like a big jump for the whole R multiverse✨can’t wait to test it out

I think ChatGPT is actually a pretty great learning tool! I still prefer Stackoverflow because I am an old man. The recent Posit session on LLMs had a nice overview of what they’re good at and integrating them into workflows

tidytuesday/May_27/May_27.ipynb at main · jacobpstein/tidytuesday A repo for posting my TidyTuesday attempts. Contribute to jacobpstein/tidytuesday development by creating an account on GitHub.

I meant to mention that I used Polars this week and found variable creation way easier than in pandas. Also, conditional, ifelse style creation is pretty smooth. Here’s my notebook if it’s of interest: github.com/jacobpstein/...

Bennedict Mathurin flies while Josh Hart tries to block him

I know it's lame to highlight a corporate-y Getty photo, but this is one of those cool basketball pics that highlights how these guys are so good at doing otherworldly stuff--like somehow shooting a ball while seemingly falling and being blocked

Love it. geom_point, quiet hero.

Yeah! It would be great to hear how other people approach these

I don’t like live coding—it’s kind of like when someone asks if you know any jokes and you can’t think of a single funny thing you’ve ever heard in your life. But it’s also probably good to live code occasionally so at least you know where you get stuck, what makes you nervous, etc.

This week's #TidyTuesday was a tough one! Lots of correlated values, no domain knowledge, and small-n groups. I spent a long time flailing around trying to figure out what might be interesting. Predicting hit points based on the other data seemed like a good way to compare model types.

Like what you did there with the ggplot theme🐲

I am also using this week’s data to kick the tires on polars! It is…fine. The .to_pandas function call is getting a lot of use

Weird, don't know why the URL didn't render as a hyperlink. Here ya go! 0196f5d5-dc61-3977-66b7-ccd1e7b9cead.share.connect.posit.cloud#/title-slide

The story I told to accompany the slides was basically about not being able to shake a stat and how even when you understand something rationally, the irrational hold it can have on you lingers

I didn't get around to doing #TidyTuesday last week because I was hustling to finish slides for a presentation to the DC Data Viz meetup. Here are the slides--https://0196f5d5-dc61-3977-66b7-ccd1e7b9cead.share.connect.posit.cloud/#/title-slide

@owenphillips.bsky.social don't quite know what to make of this, but the correlation between two point attempts and shot quality went positive on average for the first time this season. Could be spurious, could be mid-range theory at play

I did econ, but a lot of this comes from putting in reps. A new data set is like walking into a restaurant you've never been to before--you probably have a sense of where the bathroom might be, or at least what to look for, and you'll know if you've accidentally wandered into the kitchen

Totally reasonable question! It's a scatter plot, but instead of doing two continuous variables on the x and y axes, you make one of the axes factor (or character or some other class). Code is here! github.com/jacobpstein/...

thanks! I'm glad I finally took the plunge into Tidy Tuesday!

I have been re-reading Ferrante's Neapolitan Novels so this week's #TidyTuesday felt very much on theme. I started to go down a rabbit hole of spatial modeling, but decided that for getting this done while I have a little time, it's better just to make a nice descriptive plot.