iiinova — where your digital self keeps evolving

Mattan S. Ben-Shachar @mattansb.msbstats.info · 15h

To me this all feels like an offspring of "adjust for Z if it's significantly related to X and to Y" and people had to explain that sometimes you need to adjust if Z isn't related to Y at all, etc etc etc...

020

Mattan S. Ben-Shachar @mattansb.msbstats.info · 15h

I really hate this topic (sorry Peter) - adjust for what you need for causal identification and keep a mental note that adjusted vs unadjusted estimates can be quantitatively and qualitatively different for the each other?

220

Reposted by Mattan S. Ben-Shachar

CineCalendar @cinecalendar.bsky.social · 03/10/2026

Oct 3rd - It’s October 3rd. Happy #MeanGirlsDay 📽️📅 Mean Girls (2004)

2 Screenshots from the movie "Mean Girls" (2004): First, a classroom setting with a young woman with long red hair talking to a young man in a brown shirt and blue t-shirt, both sitting at desks with open books. Subtitles read, "On October third, he asked me what day it was." Second, the same students are seen from a slightly different angle; the young woman smiles slightly as she responds. Subtitles read, "It's October third."

016772

Mattan S. Ben-Shachar @mattansb.msbstats.info · 22h

The different "types" of suppression are part of the curriculum where I teach, so this paper is gonna be very useful!

110

Mattan S. Ben-Shachar @mattansb.msbstats.info · 03/10/2026

How underused would you say it is? Not at all Very |––––––––––––––––––––|

060

Reposted by Mattan S. Ben-Shachar

Hadas Weiss @hadas.bsky.social · 03/10/2026

mind if i slip into something more comfortable? (converts to ms word)

4544

Mattan S. Ben-Shachar @mattansb.msbstats.info · 03/10/2026

Screenshotting this is going to save me so much time!

1110

Reposted by Mattan S. Ben-Shachar

Dustin A. Chacón @lingdustin.bsky.social · 02/10/2026

I've heard people make claims that AI would lead to a new Renaissance, but I didn't imagine it'd happen like this:

Screenshot of a Tweet from Pope Leo XIV. It reads:

"In this era of artificial intelligence, it is becoming urgen to distinguish human art from what machines produce. There is an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation based on millions of images created by others. Algorithms lack the spark of humanity. For this reason, the Church wishes to renew an alliance with artists and cultural institutions to safeguard our humanity."

255181984803

Reposted by Mattan S. Ben-Shachar

Jordan Nafa @ajordannafa.com · 03/10/2026

The AI people co-opting and bastardizing "Bayesianism" is an unforgivable blasphemy and we ought to burn them at the stake like witches for their sins.

082

Reposted by Mattan S. Ben-Shachar

JsonGeller @jgeller1phd.bsky.social · 02/10/2026

New tutorial finally in press with @rmkubinec.bsky.social @chelseaparlett.bsky.social and @matti.vuorre.com. It has everything you could possible want: Quippy headings, a whole lot of beta, bayesian analyses and frequentist analyses side by side in harmony. link.springer.com/article/10.3...

rdcu.be

27525

Mattan S. Ben-Shachar @mattansb.msbstats.info · 01/10/2026

Me: Wait, why are you performing a cluster analysis? Student: I want to describe the data in a lower dimensional setting. Me: How many dimensions do you have? Student: 4, but the clustering seems to only be sensitive to 2. Me: And you're getting how many clusters...? Student: 3. Me: ...

1100

Mattan S. Ben-Shachar @mattansb.msbstats.info · 01/10/2026

So many people doing cluster analysis (have many hard-to-justify assumptions, hard to interpret, validate) when they really seem to just want to draw some non-parametric density plots or a couple of scatter plots with arrows on them.

151

Reposted by Mattan S. Ben-Shachar

Nicola Rennie @nrennie.bsky.social · 01/10/2026

If you missed today's @rinpharma.bsky.social workshop, here are a few of my favourite "quick fixes" for some common #ggplot2 issues 📊 1. Use {ggview} to preview plots at the same size and resolution as you want to save them with ggsave() to avoid the annoying resizing #RStats #DataViz

library(ggview)
ggplot(mtcars, aes(wt, mpg)) +
geom_point() +
canvas(
width = 6, height = 4,
units = "in",
dpi = 300
)

112741

Reposted by Mattan S. Ben-Shachar

Thomas Lin Pedersen @thomasp85.com · 30/09/2026

We have been hard at work on ggvis2 but we will have to severely restrict access to it due to safety concerns. The visualisations are just too convincing

4427

Mattan S. Ben-Shachar @mattansb.msbstats.info · 01/10/2026

I think part of the beauty is that if you know what your model specification is, it's relatively easy to "make the jump" to code it (that's why R uses the Wilkinson-Bates notation...).

110

Mattan S. Ben-Shachar @mattansb.msbstats.info · 01/10/2026

This is also why "knowing to code" means different things to different people - different levels of abstraction are required for different tasks. For fitting a GLMM I don't need to know how memory is allocated or even how different optimizers find the MLE..

110

Mattan S. Ben-Shachar @mattansb.msbstats.info · 01/10/2026

Coding is more about abstraction of a problem and how its components come together than it is about telling computers what to do. Else we would be coding our GLMMs in C or something. "Not knowing to code" is not knowing how to logically think about problems that might be solvable with a computer.

2131

Mattan S. Ben-Shachar @mattansb.msbstats.info · 30/09/2026

Me before posting: Michael is gonna say that the DGP is the joint dist., hu? 🤣

120

Mattan S. Ben-Shachar @mattansb.msbstats.info · 29/09/2026

Me ticking that box: yes, *this time* it'll work!

130

Reposted by Mattan S. Ben-Shachar

Andreas Brandmaier @brandmaier.bsky.social · 28/09/2026

Ever looked at your graph and thought: this needs a more dramatic exit strategy? Make your R base plots self-destruct with style. If you must, download here: github.com/brandmaier/d... #RStats #DataVisualization #RProgramming #DataScience #OpenSource #AcademicHumor #noAI

47622

Reposted by Mattan S. Ben-Shachar

Ian Hussey @ianhussey.mmmdata.io · 29/09/2026

Hayes has been banging this drum about idionomic analyses for 6+ years, with folk like @jamiecummins.bsky.social pointing out that its reinventing-the-wheel-but-worse for just as long. My experience is that the ACT community is insufficiently interested and aware of existing methods+stats work.

12312

Mattan S. Ben-Shachar @mattansb.msbstats.info · 29/09/2026

Of course rmANOVA only gives inference about fixed effects, and if you only want those (and not estimation or inference of random effects) GEE gives a nice easy to use shortcut.

041

Mattan S. Ben-Shachar @mattansb.msbstats.info · 29/09/2026

rmANOVA (and paired t-test) is a special case of LMMs, which means you can in theory always use LMMs. Only two advantages to rmANOVA: 1. As Cameron said, much harder to fuck up model spec. 2. With complex designs afex::aov_ez can take a fraction of the time of lme4::lmer.

130

Reposted by Mattan S. Ben-Shachar

Emily Riederer @emilyriederer.bsky.social · 28/09/2026

Yes, I would judge you for bringing iced coffee to an interview. But only because diluting previously hot brewed coffee with ice cubes is definitionally weak. Bring cold brew to show me you understand good things take time and are worth upfront investment. That’s a builder mindset

3302

Mattan S. Ben-Shachar @mattansb.msbstats.info · 28/09/2026

Okay I've never heard the tuna-piano joke and I find it very interesting that in Hebrew there's basically the same joke: What's the difference between an elephant (peel) and a piano? You can drop (lehapeel, also lit. "to the-elephant") a piano, but you can't piano an elephant.

010

Reposted by Mattan S. Ben-Shachar

Robin Blythe @rbly.bsky.social · 16/09/2026

We must pace the frontier of logistic regression. It is too powerful, and the risk to humanity of incorrect interpretations of odds ratios is an existential threat #statsky #ai

3274

Reposted by Mattan S. Ben-Shachar

Eiko Fried @eikofried.bsky.social · 15/09/2026

Because cluster analysis is always better, at least according to @dingdingpeng.the100.ci

141

Reposted by Mattan S. Ben-Shachar

Dr Mircea Zloteanu 🌺🌞🍃 @mzloteanu.bsky.social · 25/09/2026

#statstab #625 {lavinteract}: Post-Estimation Utilities for 'lavaan' Fitted Models Thoughts: "Companion toolbox for structural equation models fitted with 'lavaan'" #lavaan #sem #modelselection #modelfit #r2 #FDR #VIF #latent #rstats #r cran.r-project.org/web/packages...

cran.r-project.org

Help for package lavinteract

1122

Mattan S. Ben-Shachar @mattansb.msbstats.info · 26/09/2026

Wow this is great! There is also some support for lavaan in emmeans via the semTools package (written by your truly) cran.r-project.org/web/packages... Would be interesting to try and apply this with marginaleffects.

cran.r-project.org

Help for package semTools

010

Mattan S. Ben-Shachar @mattansb.msbstats.info · 26/09/2026

Had to defend my thesis proposal. After the break all of the panelists made their comments. One of them commented on the cookies I had brought being store-bought and not home-baked. That was all he said. The university has since banned the practice of students providing snacks and refreshments.

150

Mattan S. Ben-Shachar @mattansb.msbstats.info · 25/09/2026

I'm not well versed in rlang. How would you do it?

000

Mattan S. Ben-Shachar @mattansb.msbstats.info · 25/09/2026

Yes, I say things along those lines - use LLMs for revision, writing practice questions (but not summarizing or answering/solving questions).

010

Reposted by Mattan S. Ben-Shachar

Hadas Weiss @hadas.bsky.social · 25/09/2026

ending my cover letter with “if you even care”

410614

Mattan S. Ben-Shachar @mattansb.msbstats.info · 25/09/2026

Too true. How many errors in foo(x) are resolved by foo(as.data.frame(x))...

011

Mattan S. Ben-Shachar @mattansb.msbstats.info · 25/09/2026

Students are well within their rights to work towards failure. I don't force tidymodels (I even accept sklearn).

010

Mattan S. Ben-Shachar @mattansb.msbstats.info · 24/09/2026

Sure. But... do you like it enough to write code you didn't learn to properly use and lose 20 points because of that?

140

Mattan S. Ben-Shachar @mattansb.msbstats.info · 24/09/2026

I mean I know some people dislike tibbles, but this seems a bit much...

210

Mattan S. Ben-Shachar @mattansb.msbstats.info · 24/09/2026

Did they manage to avoid data leackage in their model training? No. But did they properly pre-process the data? Also no.

2192

Mattan S. Ben-Shachar @mattansb.msbstats.info · 24/09/2026

Me: teaches a full semester of machine learning with #tidymodels. Student: I don't like tibbles, so for the final assignment I decided to code everything in base #rstats.

static.klipy.com

Bart Simpson: What an Odd Thing to Say

ALT: Bart Simpson: What an Odd Thing to Say

5201

Reposted by Mattan S. Ben-Shachar

Alex Kale @kalealex.bsky.social · 24/09/2026

What AI scales about science are the parts that are most broken and gameable. With the exception of combinatorial problems like protein folding, we need more thinking and less rote computation.

0236

Mattan S. Ben-Shachar @mattansb.msbstats.info · 24/09/2026

It already is - uses vctrs::vec_if_else under the hood.

110

Reposted by Mattan S. Ben-Shachar

CineCalendar @cinecalendar.bsky.social · 24/09/2026

Sep 24th - East German Nazis will kill Dr. Paul Flammond's daughter, unless he completes the Polaris Naval Mine by this date. 📽️📅 Top Secret! (1984)

3 Screenshots from the movie "Top Secret!" (1984): First, an older man in a white lab coat and glasses speaks urgently, with the subtitle reading “They’ll kill her unless I complete the Polaris Mine by Sunday.” Second, the man gestures toward a wall calendar labeled “ACME LAB EQUIPMENT” showing a pin-up style image and the month “Septhauzen,” pointing to Sunday the 24th. Third, a young man in a blue shirt responds in confusion, with the subtitle reading “Sunday? That’s Simhath Torah.”

36313