Arthur Katossky
LightNews LightNews
katossky.bsky.social
Arthur Katossky
@katossky.bsky.social
150 followers 510 following 120 posts
Des maths, de l'informatique et un twist de sciences sociales
Posts Replies Media Videos
Reposted by Arthur Katossky
arindube.bsky.social
When states like Colorado passed policies requiring employers to disclose salary information in job postings, what happened?

It increased competition, and raised wages, without harming employment or changing skill requirements.

Improved functioning of markets, helped workers.
November 17, 2025 at 4:41 PM
Reposted by Arthur Katossky
datavisfriendly.bsky.social
#TodayinHistory #dataviz #OTD 📊
🎂Nov 11, 1851 Jacques Bertillon born in Paris, France 🇫🇷

1896: Use of area rectangles on a #map to display two variables and their product (population of arrondisements in Paris, percent foreigners; area = absolute number of foreigners)
Bertillion's map of Paris, titled Les etrangers a Paris en 1891. In each arrondisement, there is a black rectangle with width ~ population, heighr ~ percent foreigners. So, the area represents absolute number of foreigners. Photo of Jacques Bertillon
November 12, 2025 at 1:05 AM
Reposted by Arthur Katossky
trfetzer.com
Today in economic complexity -- or, how we can use some techniques we are building to revolutionize how we can think about modern industrial policy. This time: how can animal bones serve as an input into vacuum pumps? There are many paths along a knowledge graph.
November 7, 2025 at 3:55 AM
Reposted by Arthur Katossky
luok.ai
Toyota just demoed a personal mobility chair that literally walks—up stairs, sideways, and into a car.

Meet “walk me”: a four-legged, stair-climbing, height-adjustable mobility device aimed at seniors and people with limited mobility.
November 5, 2025 at 4:53 PM
Reposted by Arthur Katossky
frodsan.bsky.social
A single call to function graph2ppt or graph2doc from the {export} package will produce fully editable graphs in PowerPoint and Word, respectively. Easy! Works with base, ggplot...

cran.r-project.org/web/packages... #rstats
export: Streamlined Export of Graphs and Data Tables
Easily export 'R' graphs and statistical output to 'Microsoft Office' / 'LibreOffice', 'Latex' and 'HTML' Documents, using sensible defaults that result in publication-quality output with simple, stra...
cran.r-project.org
November 5, 2025 at 10:52 AM
Reposted by Arthur Katossky
mattansb.msbstats.info
How to fit a model without estimating any parameters?

Use a formula with
y ~ 0 + offset(...)
This works with any formula interface, and with #tidymodels this can be passed to add_model().

#rstats
November 5, 2025 at 10:56 AM
Reposted by Arthur Katossky
danpsimpson.bsky.social
Distance uphill and a distance downhill should be measured in asymmetric effort.
November 5, 2025 at 6:01 AM
Reposted by Arthur Katossky
pjanuel.bsky.social
L'an dernier, les dépenses de recours aux cabinets conseil ont augmenté de 8% (et pas 31%, comme l'affirment ceux qui confondent AE et CP).
Si on compare à 2022, la baisse est de 40%. Mais les cabinets fontt de la résistance dans certains ministères et directions
www.lalettre.fr/fr/politique...
Cabinets de conseil : la direction de la haute fonction publique peine à freiner ses dépenses - LA LETTRE
Si le recours par l'État aux cabinets de conseil a fortement reculé depuis l'affaire McKinsey, certains ministères et les services du premier ministre continuent toutefois de solliciter les conseils
www.lalettre.fr
October 29, 2025 at 6:23 AM
Reposted by Arthur Katossky
vandeneeckhoutkoen.bsky.social
Qualitative information design: don't underestimate how much you can design with only text!

1/8
A “Before and After” comparison of a research structure diagram. Before: A dense yellow-orange block diagram with boxes and sub-boxes showing sections such as “Introduction and Motivation,” “The Concept of Uncertainty,” “Uncertainty Representation Approaches,” “Recent Advances in UnViz,” “Evaluation of UnViz Techniques,” and “Future Research Directions.” Subtopics like “Data Sampling Uncertainty,” “Decision Uncertainty,” “Machine Learning,” “Big Data,” and “User Studies” are clustered, making the layout busy and visually cluttered. After: A clean, vertical flow diagram resembling a timeline or hierarchy, numbered 1–5. Each main section—like “The concept of uncertainty,” “Uncertainty representation approaches,” “Evaluation of uncertainty visualization techniques,” “Recent advances in uncertainty visualization,” and “Future research directions”—branches into smaller indented topics with light connecting lines. A gray line runs through the sequence toward “Conclusion.” New elements are marked with an exclamation symbol. The redesign replaces the crowded box layout with an elegant, linear structure that clarifies relationships and sequence.
October 28, 2025 at 2:26 PM
Reposted by Arthur Katossky
leecrawfurd.bsky.social
Vietnam is a massive positive outlier on international school tests riseprogramme.org/node/141.html

Striking that its also a large outlier in how valued teachers feel by both parents and policymakers

(From the latest TALIS survey published earlier this month
www.oecd.org/en/publicati...)
October 22, 2025 at 10:28 PM
Reposted by Arthur Katossky
undercoverhist.bsky.social
Dozens of new ocean methane seeps have recently been discovered. This is not factored in current climate scenarios

www.nature.com/articles/s41...
October 20, 2025 at 7:51 PM
Reposted by Arthur Katossky
guillaume.champeau.info
Cette stat est folle : "61% des usagers rencontrent des difficultés avec leurs démarches administratives, contre 39% il y a 8 ans"

On voit là l'effet de la numérisation et du manque de considération assez sérieuse pour l'illectronisme total ou partiel. www.bfmtv.com/economie/eco...
De plus en plus d'usagers galèrent dans leurs démarches administratives (61% contre 39% il y a 8 ans) et un quart abandonne
61% des usagers rencontrent des difficultés avec leurs démarches administratives, contre 39% il y a 8 ans, selon une enquête du Défenseur des droits (DDD). L'étude pointe des "fractures persistantes" ...
www.bfmtv.com
October 13, 2025 at 1:12 PM
Reposted by Arthur Katossky
gnoblet.bsky.social
One of my graph has just entered the "best" section of the r-graph-gallery.com with a tutorial.

Featuring a waffle chart for time series, where the subtitle serves as a colorful legend.

Thanks a lot to @yan-holtz.bsky.social and @soeundataviz.bsky.social for adding it! #rstats #dataviz #ggplot2
October 10, 2025 at 6:24 AM
Reposted by Arthur Katossky
wombat.numbat.space
Cannot post. Too busy doing ggplot battles.

#RStats #WOMBAT2025
www.ggplotbattles.dev/challenges/v...
Screenshot of a website showing a ggplot2 plot that you need to replicate with code written on a text box (code blurred! no cheating!)
September 30, 2025 at 2:04 AM
Reposted by Arthur Katossky
mjskay.com
fun! are we allowed to improve the chart?

that density estimator doesn't account for the boundedness of group C and it's usually possible to do better than random jitter
density + 50% and 95% intervals + mean + compressed dotplot with exact dot positions on three groups, including group C which is automatically estimated using a bounded KDE due to a lower bound. library(ggplot2) library(ggdist) spreads |> ggplot(aes(Group, Value, fill = Group)) + geom_weave( side = "left", scale = 0.75, alpha = 0.35, binwidth = unit(1, "mm"), overlaps = "keep", color = NA ) + stat_slab( scale = 0.75, alpha = 0.65, color = "black", linewidth = 0.25 ) + stat_pointinterval( fill = "white", shape = 23, point_interval = "mean_qi", .width = c(.5, .95) ) + guides(fill = "none") + labs(x = NULL) + theme_minimal(base_size = 13) + theme(panel.grid.minor = element_blank())
September 30, 2025 at 4:15 AM
Reposted by Arthur Katossky
evegwood.com
i can't reply to the original post, but potentially this? Alien #5 by Phillip Kennedy Johnson and Julius Ohta @jongraywb.bsky.social
September 26, 2025 at 9:50 AM
Reposted by Arthur Katossky
transport-talk.bsky.social
From my #rstats bookmarks:
{flow} - Visualize functions and scripts as flow diagrams: github.com/moodymudskip...
September 24, 2025 at 2:52 AM
Reposted by Arthur Katossky
p-hunermund.com
This new CESifo working paper on covariate balancing looks very interesting. @jmwooldridge.bsky.social

Link: www.ifo.de/DocDL/cesifo...
September 23, 2025 at 9:06 AM
Reposted by Arthur Katossky
vincentab.bsky.social
Whoa—my book is up for pre-order!

𝐌𝐨𝐝𝐞𝐥 𝐭𝐨 𝐌𝐞𝐚𝐧𝐢𝐧𝐠: 𝐇𝐨𝐰 𝐭𝐨 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭 𝐒𝐭𝐚𝐭 & 𝐌𝐋 𝐌𝐨𝐝𝐞𝐥𝐬 𝐢𝐧 #Rstats 𝐚𝐧𝐝 #PyData

The book presents an ultra-simple and powerful workflow to make sense of ± any model you fit

The web version will stay free forever and my proceeds go to charity.

tinyurl.com/4fk56fc8
September 17, 2025 at 7:49 PM
Reposted by Arthur Katossky
gallegre.bsky.social
Faut-il réduire l'enseignement supérieur ?
C'est le programme de l'ami Julien Damon. Au moment où je m'en offusque, j'apprends que c'est aussi le programme de Renaissance, dans une note programmatique intitulée "le camp de l'espoir" (!)
September 17, 2025 at 6:35 AM
Reposted by Arthur Katossky
louisbarclay.bsky.social
Q. Who aligns the aligners?
A. alignmentalignment.ai

Today I’m humbled to announce an epoch-defining event: the launch of the 𝗖𝗲𝗻𝘁𝗲𝗿 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗼𝗳 𝗔𝗜 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗖𝗲𝗻𝘁𝗲𝗿𝘀.
Center for the Alignment of AI Alignment Centers
We align the aligners
alignmentalignment.ai
September 11, 2025 at 1:17 PM
Reposted by Arthur Katossky
gmcd.bsky.social
To borrow another example, taken from the `dbreg` README: github.com/grantmcdermo...

Here I am running a fixed-effects regression on 180 million(!) row parquet dataset... and it completes **< 2 seconds**... on my laptop 🤯

This is powered by @duckdb.org under the hood.

#rstats #econsky
Running dbreg::dbreg() on a full year of NYC taxi data... and it takes less than 2 seconds. dbreg( tip_amount ~ fare_amount + passenger_count | month + vendor_name, path = "read_parquet('nyc-taxi/**/*.parquet')", ## path to hive-partitioned dataset vcov = "hc1" ) #> [dbreg] Estimating compression ratio... #> [dbreg] Data has 178,544,324 rows and 24 unique FE groups. #> [dbreg] Using strategy: compress #> [dbreg] Executing compress strategy SQL #> #> Compressed OLS estimation, Dep. Var.: tip_amount #> Observations.: 178,544,324 (original) | 70,782 (compressed) #> Standard Errors: Heteroskedasticity-robust #> Estimate Std. Error t value Pr(>|t|) #> fare_amount 0.106744 0.000068 1564.742 < 2.2e-16 *** #> passenger_count -0.029086 0.000106 -273.866 < 2.2e-16 *** #> --- #> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
August 25, 2025 at 6:38 PM
Reposted by Arthur Katossky
dingdingpeng.the100.ci
Ever stared at a table of regression coefficients & wondered what you're doing with your life?

Very excited to share this gentle introduction to another way of making sense of statistical models (w @vincentab.bsky.social)
Preprint: doi.org/10.31234/osf...
Website: j-rohrer.github.io/marginal-psy...
Models as Prediction Machines: How to Convert Confusing Coefficients into Clear Quantities Abstract Psychological researchers usually make sense of regression models by interpreting coefficient estimates directly. This works well enough for simple linear models, but is more challenging for more complex models with, for example, categorical variables, interactions, non-linearities, and hierarchical structures. Here, we introduce an alternative approach to making sense of statistical models. The central idea is to abstract away from the mechanics of estimation, and to treat models as “counterfactual prediction machines,” which are subsequently queried to estimate quantities and conduct tests that matter substantively. This workflow is model-agnostic; it can be applied in a consistent fashion to draw causal or descriptive inference from a wide range of models. We illustrate how to implement this workflow with the marginaleffects package, which supports over 100 different classes of models in R and Python, and present two worked examples. These examples show how the workflow can be applied across designs (e.g., observational study, randomized experiment) to answer different research questions (e.g., associations, causal effects, effect heterogeneity) while facing various challenges (e.g., controlling for confounders in a flexible manner, modelling ordinal outcomes, and interpreting non-linear models). Figure illustrating model predictions. On the X-axis the predictor, annual gross income in Euro. On the Y-axis the outcome, predicted life satisfaction. A solid line marks the curve of predictions on which individual data points are marked as model-implied outcomes at incomes of interest. Comparing two such predictions gives us a comparison. We can also fit a tangent to the line of predictions, which illustrates the slope at any given point of the curve. A figure illustrating various ways to include age as a predictor in a model. On the x-axis age (predictor), on the y-axis the outcome (model-implied importance of friends, including confidence intervals). Illustrated are 1. age as a categorical predictor, resultings in the predictions bouncing around a lot with wide confidence intervals 2. age as a linear predictor, which forces a straight line through the data points that has a very tight confidence band and 3. age splines, which lies somewhere in between as it smoothly follows the data but has more uncertainty than the straight line.
August 25, 2025 at 11:49 AM
Reposted by Arthur Katossky
blasbenito.com
In case you didn't know:

You can run your package unit tests with {testthat} in parallel with two simple steps (see testthat.r-lib.org/articles/par...):

tldr:
1. Add `Config/testthat/parallel: true` to DESCRIPTION.
2. Add `TESTTHAT_CPUS=8` to your .Renviron and restart R.

#rstats
Running testthat unit tests in parallel with Rstudio.
August 18, 2025 at 7:24 AM
Reposted by Arthur Katossky
patrickaubert.bsky.social
A ce sujet, extrait de l'étude d'impact du système universel de retraite de 2019 :
August 20, 2025 at 6:37 AM