Content Warning

I guess now is a good time to start posting more on Mastodon, and provide a (re)introduction. I will mostly post on the #rstats tag, and I believe it is awesome if the community can find a place to stay on here. I am a researcher at Heinrich Heine University D眉sseldorf at the Department of Psychology. I am interested in programming, statistics and optimization. My most dear #rstats project is the anticlust package (https://github.com/m-Py/anticlust), which implements the method of anticlustering.
+ -

Content Warning

Anticlustering is a method to divide a set of objects into groups, while ensuring that the different groups are similar to each other. One example would be to divide a cohort of students into school classes, and the different classes should be similar on demographic composition, gender, and maybe previous grades. Ever since I implemented the openly available anticlust #rstats package, I realized that there is a wide range of different applications, which I had not thought of before.

Content Warning

Yesterday, we posted a new preprint on the anticlustering method on bioRxiv, where we developed the the must-link feature for anticlustering (https://doi.org/10.1101/2025.03.03.641320). It can be used to ensure that certain objects are assigned to the same group. Sticking with the school example, we might strive for similarity between classes regarding composition, gender, and previous grades, but we might assign certain cliques of friends to the same class through "must-link constraints".