iiinova — where your digital self keeps evolving

Reposted by Daniel 🕹️

Resiljens @stadioncheck.de · 02/10/2026

Altona-Fans mit wichtiger Message. #cornell57

Fußballfans halten ein Banner hoch: „Not all men, but always a man #Cornell57 Schaut nicht weg!“

29821

Reposted by Daniel 🕹️

Grant McDermott @gmcd.bsky.social · 28/09/2026

𝐭𝐢𝐧𝐲𝐩𝐥𝐨𝐭 v0.8.0 now available on CRAN 🎉 grantmcdermott.com/tinyplot/ This is another big release: new plot types, finer facet & axis control, smarter handling of categorical data, and lots more. Same tiny footprint, though. Short highlight thread below 🧵 #rstats

Heatmap of the mtcars dataset, scaled within each column. Code: https://github.com/grantmcdermott/tinyplot/blob/main/vignettes/gallery_figs/heatmap-mtcars.R

Histogram of New York temperatures, with bars coloured by their own x values. Code: plt(~ Temp | Temp, data = airquality, type = "hist", theme = "classic")

Hexbin plot of 20,000 bivariate normal draws. Code: https://github.com/grantmcdermott/tinyplot/blob/main/vignettes/gallery_figs/hexbin.R

Stacked area plot of telephones in use by world region, 1956-1961. Code: plt(WorldPhones[-1, ], type = type_area(stack = TRUE, byord = "end"))

13513

Daniel 🕹️ @strengejacke.de · 28/09/2026

As long as they don't go demon hunter or death knight, it's fine

010

Reposted by Daniel 🕹️

Katharina Nocun @kattascha.bsky.social · 27/09/2026

Hm, ich weiß, es ist ein beliebter Take von US-Rechten, liberalen Frauen Intoleranz vorzuwerfen, weil sie seltener rechte Männer daten wollen. Ergo tragen sie Schuld an männl. Einsamkeit & Radikalisierung, usw. Aber müssen wir wirklich jeden Quatsch hierzulande kopieren? Es langweilt mich. 🙄

541828311

Daniel 🕹️ @strengejacke.de · 24/09/2026

First R lesson: avoid tibbles wherever you can. 😎 It can save you years of searching for an error in your code when there's actually none... tidyverse with data frames would be nice.

170

Daniel 🕹️ @strengejacke.de · 23/09/2026

Or: can they be tested on CRAN? Sometimes, you want that test to work on CRAN, that might be an exception for snapshot tests

100

Daniel 🕹️ @strengejacke.de · 23/09/2026

Are snapshots tested on CRAN?

100

Daniel 🕹️ @strengejacke.de · 21/09/2026

"Intersubjektivität" presupposes subjects, and presupposes a 'connection' (meta-subject?) between these subjects that does not exist. See Luhmann, Die Wissenschaft der Gesellschaft - recommended reading, even if you don't want to follow that point of view ;-)

010

Daniel 🕹️ @strengejacke.de · 19/09/2026

So you need some kind of randomisation?

110

Daniel 🕹️ @strengejacke.de · 18/09/2026

Is Regression to the Mean a thing in longitudinal analysis using mixed models with random slopes (and interaction term between time x group)? And if so, to which extent? Or can we be relaxed due to shrinkage, and if we use pseudo-randomization via counterfactual predictions?

231

Daniel 🕹️ @strengejacke.de · 17/09/2026

The guillotine for the garbage-can regression...

030

Daniel 🕹️ @strengejacke.de · 09/09/2026

Nice to hear! 🤩 If you're looking for some resources, we have slides here: easystats.github.io/easystats/ar... And of course lot's of working examples and detailed introductions here: easystats.github.io/modelbased/ But I agree with Vincent, often, the "simple use cases" are sufficient for most.

easystats.github.io

Learning resources

001

Daniel 🕹️ @strengejacke.de · 22/08/2026

I was attending the 21st @eshms.bsky.social conference and beyond the many great sessions there was also an amazing session about the quantitative #MAIHDA approach. I updated the #rstats #easystats tutorial on MAIHDA modelling easystats.github.io/modelbased/a..., including some new things I learned.

easystats.github.io

Case Study: Intersectionality Analysis Using The MAIHDA Framework

030

Reposted by Daniel 🕹️

Aki Vehtari @avehtari.bsky.social · 13/08/2026

All four books I've co-authored are freely available online for non-commercial use: - Bayesian Workflow at avehtari.github.io/Bayesian-Wor... Links to other three books are in the quoted post 👇 (too many books to fit in one post!)

avehtari.github.io

Bayesian Workflow book: Website – Bayesian Workflow book

Website for the Bayesian Workflow book by Gelman, Vehtari, McElreath, et al. — case studies, code, and exercises in R and Stan.

8283134

Daniel 🕹️ @strengejacke.de · 05/08/2026

Still, "evidence" does not indicate "clinical importance". You should take these two into account (not a new insight, I know...).

130

Daniel 🕹️ @strengejacke.de · 05/08/2026

Just for the record, this is a good example for the `post_process` argument in {modelbased}...

library(modelbased)
data(mtcars)
m <- glm(am ~ vs, family = binomial, data = mtcars)

estimate_means(
m,
"vs",
transform = effectsize::probs_to_odds,
predict = "response",
post_process = difference ~ pairwise
)
#> Post-processing difference ~ pairwise...
#> Estimated Marginal Means
#>
#> Parameter | Probability | SE | 95% CI | z
#> ------------------------------------------------------
#> (b2) - (b1) | 0.50 | 0.17 | [0.16, 0.84] | 2.88
#>
#> Variable predicted: am
#> Predictors modulated: vs
#> Predictions are on the response-scale.

030

Daniel 🕹️ @strengejacke.de · 05/08/2026

"The p-*value* is not binary" - not an exact definition, but a good starting point.

010

Daniel 🕹️ @strengejacke.de · 28/07/2026

Which nowadays means "Claude or it didn't happen", with no chance to validate the output.

000

Daniel 🕹️ @strengejacke.de · 27/07/2026

Sadly, no. That's pretty much the issue.

000

Daniel 🕹️ @strengejacke.de · 19/07/2026

I hope US Americans don't think the World Cup final represents the real spirit of football in any way.

000

Daniel 🕹️ @strengejacke.de · 08/07/2026

You can disable extensions from being re-installed, though.

030

Daniel 🕹️ @strengejacke.de · 08/07/2026

I think you don't need a subscription for GitHub NES (next edit suggestion) to work. But I agree, the GitHub NES is everything but "discreet". I uninstalled the posit AI extension, since it didn't work anyway (except the annoying GitHub NES).

130

Daniel 🕹️ @strengejacke.de · 08/07/2026

Here's an updated vignette for the modelbased package, which is probably clearer in its presentation and therefore the concepts are maybe easier to understand. #easystats #rstats easystats.github.io/modelbased/a...

easystats.github.io

Understanding marginalization methods

152

Daniel 🕹️ @strengejacke.de · 29/06/2026

The best thing about the football game #GERPAR are the players singing the national anthems 😂🪇🙉

000

Daniel 🕹️ @strengejacke.de · 27/06/2026

Only payable with this plectrum

000

Daniel 🕹️ @strengejacke.de · 24/06/2026

All material (code, data) is freely available at osf.io/8evy5/, but you can also use the easy-to-use function `check_priors()` in the #easystats {performance}📦 to conduct prior predictive checks: easystats.github.io/performance/...

easystats.github.io

Prior predictive checks — check_priors

Simulates from the prior marginal distribution of the data to assess the consistency of the chosen priors with domain knowledge (Gabry et al. 2019, Lüdecke et al. 2026) and creates a visualization fro...

050

Daniel 🕹️ @strengejacke.de · 24/06/2026

Bayesian priors aren't just arbitrary guesses - you can (and should) validate them. Our paper shows how to use prior predictive checks to map your domain knowledge onto the model, ensuring your assumptions generate realistic, well-calibrated priors. #Rstats #Stan #Bayes

151

Daniel 🕹️ @strengejacke.de · 24/06/2026

Subjectivity concerns holding you back from Bayesian methods? Our tutorial walks through a real-world case study using "believer", "agnostic", and "skeptical" priors to run a built-in sensitivity analysis - ensuring robust and transparent science >

150

Daniel 🕹️ @strengejacke.de · 24/06/2026

Small samples or rare events making your frequentist models unstable? In our paper, we show how Bayesian informative priors act as a regularizing force, narrowing credible intervals and preventing implausible estimates when data is sparse. >

130

Daniel 🕹️ @strengejacke.de · 24/06/2026

A new paper by @dominiquemakowski.bsky.social, @mattansb.msbstats.info, me and colleagues just out! We show how to choose informative priors in Bayesian regression models using a systematic simulation study and a practical step-by-step tutorial in #Rstats and #Stan! doi.org/10.3389/fpsy... >

17623

Daniel 🕹️ @strengejacke.de · 15/06/2026

Not a book, but a monthly magazine. Listings were available in Basic and ASM language.

020

Daniel 🕹️ @strengejacke.de · 15/06/2026

Fourth explanation: it's football. Hard to predict the outcome.

030

Daniel 🕹️ @strengejacke.de · 14/06/2026

Curaçao fans were in a fantastic mood even after their defeat against Germany—and then, the German Ballermann music started playing in Houston 🙈 Horrible...

000

Daniel 🕹️ @strengejacke.de · 14/06/2026

It's a FIFA/US event, you can be happy to find some small bits of football between all ads and shows 😜

020

Daniel 🕹️ @strengejacke.de · 24/05/2026

And there's a package vignette with some more details: easystats.github.io/performance/...

easystats.github.io

Checking model assumption - linear models

041

Reposted by Daniel 🕹️

RASH und Subkultur @rashundsubkultur.bsky.social · 20/05/2026

Sehr nützliche Übersicht über antifaschistische Rechercheseiten und andere linke Seiten:

de.indymedia.org

Das Rad muss nicht neu erfunden werden 2.0 / Link-Liste | de.indymedia.org

0167

Daniel 🕹️ @strengejacke.de · 21/05/2026

The biggest hurdle in switching from SPSS to R is thinking in programming logic. Instead of single datasets or temporary outputs, R treats everything as an object. To master this mindset, I highly recommend Andy’s video series - especially "Creating an Object" and "What is a function?" #rstats

040

Daniel 🕹️ @strengejacke.de · 20/05/2026

Even if you don't like Bayes factors, if you read this, you'll like them! #easystats #rstats #bayestestR

050

Daniel 🕹️ @strengejacke.de · 20/05/2026

Good to hear, I thought I missed something... And yes, for this particular example, I was not looking for counterfactual predictions.

020

Daniel 🕹️ @strengejacke.de · 20/05/2026

Still not get it working 🙈 Can you rewrite this code with a different hypothesis argument and make it return the same results? (code in alt text)

set.seed(123)
n <- 200
d <- data.frame(
outcome = rnorm(n),
grp = as.factor(sample(c("treatment", "control"), n, TRUE)),
episode = as.factor(sample(1:3, n, TRUE)),
sex = as.factor(sample(c("female", "male"), n, TRUE, prob = c(0.4, 0.6)))
)
model2 <- lm(outcome ~ grp * episode, data = d)

avg_predictions(
model2,
by = c("episode", "grp"),
hypothesis = "(b1 - b3) = (b2 - b4)"
)

100

Daniel 🕹️ @strengejacke.de · 20/05/2026

Does difference ~ pairwise also return difference-in-difference-contrasts (interaction contrasts)? Couldn't reproduce the above results yet that way.

200

Daniel 🕹️ @strengejacke.de · 20/05/2026

We put quite some effort into extracting the related labels to "b1", "b2" etc., to put that information for users in the footer, so you can easily see whether you chose the correct coefficients to compare.

130

Daniel 🕹️ @strengejacke.de · 20/05/2026

Code in Alt-text... You always have to look at the table of predictions, to find out the b-terms/rows. Since the order of coefficients is different between modelbased and marginaleffects, you see different values for the hypothesis-argument here.

library(modelbased)
library(marginaleffects)

set.seed(123)
n <- 200
d <- data.frame(
outcome = rnorm(n),
grp = as.factor(sample(c("treatment", "control"), n, TRUE)),
episode = as.factor(sample(1:3, n, TRUE)),
sex = as.factor(sample(c("female", "male"), n, TRUE, prob = c(0.4, 0.6)))
)
model2 <- lm(outcome ~ grp * episode, data = d)

estimate_contrasts(model2, c("episode", "grp"), comparison = "(b1 - b2) = (b4 - b5)", estimate = "average")

avg_predictions(
model2,
by = c("episode", "grp"),
hypothesis = "(b1 - b3) = (b2 - b4)"
)

030

Daniel 🕹️ @strengejacke.de · 20/05/2026

Here's how to do it with #easystats modelbased, which internally uses marginaleffects: easystats.github.io/modelbased/a... Let me look into the codebase, then I can tell you the marginaleffects code later,too

easystats.github.io

Contrasts and pairwise comparisons

121