Jean-Baptiste

A tale of two scripts: Applying the principle of least complexity to simplified and traditional Chinese - Journal of Cultural Cognitive Science Is there a relationship between the frequency of characters and their respective complexity? This hypothesis, which we may call the Principle of Least Complexity, was recently tested across a number o...

Are frequency and complexity related in Chinese characters/words? Our analysis shows the association is negative but very weak. The article, pre-registration, and data are openly available to whoever wants to read about it. Feel free to drop a comment if you have questions.
doi.org/10.1007/s418...

Hey #stats and #methods people!

I will be teaching #psych MA students statistical modelling this year.

I'm looking for podcast *episode* recommendations 🎧

Give me your best!

(See list of topics I will be introducing below:)

I was going to write a blog post about this, but I don't think there is enough material to justify this.

I've used the term "Science Support Staff" pejoratively to refer to applied statistics and data science. Obviously, we need science support staff, but the role can grind one down easily.

Republicans slam Trump’s firing of Bureau of Labor Statistics chief President claimed that recently released figures showing jobs slowdown were ‘rigged’ to make him look bad

Stats matter. A lot. www.theguardian.com/us-news/2025...

the viz *suggested by OWID ‒ sorry for the typo

I find that a before-after slopegraph would probably have been a better visualization for these data (see Tufte www.edwardtufte.com/notebook/slo...). With the viz suggest by OWID, I have to go back and forth to understand how the ranking changes, while this is quicker to see with a slopegraph.

On notera par contre qu'il n'y aucun démenti sur le fait de frapper les murs gratuitement. Plus sérieusement, bon rétablissement, c'est dans ces moments-là qu'on se rend compte de l'utilité des pouces dans la vie quotidienne !

Anyone switched from #RStudio to #Positron for teaching? I'm moving to #Positron in my own usage and wondering if I should switch for my Data Analysis class. We only code in #RStats, but maybe #Positron is a long-term better tool for students to learn? Appreciate any thoughts/pointers to materials.

In any case, I think it's important to make them aware that there are alternatives to the specific tool you are using in your teaching. Let's not make people too dependent on one organization or company, this is not good for anyone long term.

Any tool can become obsolete or undesirable at some point, for a variety of reasons. I think tools should be taught as much as possible in an agnostic manner, i.e. here making students able to switch without too much complications to an IDE different from the one they were taught initially.

Subsite Passwords / Basic Authentication in Quarto · quarto-dev quarto-cli · Discussion #7511 Description I'm creating a website project in quarto and I want to implement basic authentication. Right now I have a page index.qmd with some public content and a button that links to another page...

this related discussion might be informative: github.com/quarto-dev/q...

Étude inédite de @Insee.fr : mesurer la mobilité sociale de parents à enfants en incluant dans l’analyse le statut d’emploi (indépendant ou salarié). À lire ! (1/7)
insee.fr/fr/statistiq...

Using a small multiple of choropleth maps showing percentages may be useful to avoid resorting to pie charts, which may be hard to read and interpret (e.g. on this specific map, small proportions may be hard to spot in areas with small circles) en.wikipedia.org/wiki/Small_multiple

How flawed data is leaving the UK in the dark Trouble at the Office for National Statistics means policymakers cannot rely on key economic numbers, including on employment

Big piece on how the UK is not getting the data it needs and how this is affecting policymaking. A key issue seems to be that the shift from surveys to admin data has not worked well as govt departments have not allowed the ONS access to the data it needs
on.ft.com/3WSCN6h

Ah, I see the problem, hard to solve.

My first impulse was to interpret colors as showing the degree of agreement, when it actually informs about geographic categories. Not intuitive at all.

Quite an odd choice of not using a map for this visualization.

No, I don't mean that! Really, they organize events targeted at everyone interested in the role of statistics in public issues.

One of the two authors gives regular paid masterclasses called "Take Control of Your Health with Science-Backed Insights"...

(well, this is passed the first impression that there's no data at all on this graph)

And are we supposed to compare the four graphs? I don't see how, as on the last graph, the scale of the vertical axis is multiplied by 10 compared to the others -e.g. it may hide some important variations compared to the three other graphs.

De rien. Je me souviens que j'avais été tout aussi perplexe face à leurs données ouvertes sur le nombre d'inscriptions (où ils utilisent le même système où une ligne = plusieurs personnes). Ça m'avait pris un petit moment d'exploration de leur documentation pour comprendre d'où venait le problème.

Ce qui est affiché sur la page web du ministère sous la colonne "filtres" n'est pas le nombre de diplômes, mais le nombre de lignes dans le fichier xlsx ou csv. Or, chaque ligne du fichier représente plusieurs délivrances de diplômes (et non une seule), donc le chiffre est bien plus grand que 90000.

The French Statistical Society regularly organizes events targeted at laypeople, to make stats more accessible. I guess there are similar things in other countries.

This is vital in democracies, it's impossible to take informed political decisions otherwise.

Another drawback is its slowness with heavy computation. I stumbled upon the issue when trying to solve a triple integral numerically for computing Bayesian assurance. I didn't try with R yet, but I suspect the problem will be the same and that I'll have to turn to a lower-level language eventually.

Attention d'ailleurs, il y a une autre colonne intitulée "Nombre de diplômes délivrés", qui donne un nombre moins important (592.332 pour 2021-2022). La différence entre ces deux chiffres vient des "diplômes intermédiaires" délivrés (225.601).

C'était pour quelle(s) année(s) ? De mon côté, avec un calcul rapide dans Excel, je trouve 817.933 diplômes délivrés au total pour l'année universitaire 2021-2022 (somme faite sur la colonne "Nombre total de diplômes délivrés").

Souvent dans les données ouvertes du ministère, chaque ligne ne représente pas une observation ou un individu, mais plusieurs. Je ne sais pas si c'est l'origine du problème ici, mais si c'est le cas, il faut faire le calcul sur une colonne qui devrait s'appeler "total" ou "Nombre de diplômes".

je n'avais pas vu la discussion que ça avait généré ici bsky.app/profile/pecq... , ça répond à ma question.