@majavonwestphal Da spielt dann die mangelnde Dokumentation rein. Das ist auch ein problem, aber gegeben der guten Replizierbarkeit auch nichts, was die Forschung generell in Frage stellt.
@majavonwestphal Daten und oder Code werden häufig nicht geteilt. Daher das schlechte abschneiden. Bei der Replizierbarkeit steht Bildungsforschung recht gut da.
Yep, the most underappreciated lesson from the replication crisis is that good science doesn’t boil down to easily automatable criteria.
Maxing out transparency & robustness may improve status quo, but epistemic virtues of great science are not reducible to the right workflow.
After 5 yrs of data collection, our WARN-D machine learning competition to forecast depression onset is now LIVE! We hope many of you will participate.
If you share a single thing of my lab this year, please make it this competition.
https://t.co/BeYDExi6qZ
New article: Examining the relationship between disciplinary counterspaces for undergraduate women and sense of belonging: A physics exemplar
Check it out here https://t.co/YcNe0Nw5Ul
Agree. ChatGPT is most efficient and least harmful when you already have enough knowledge and resources to fact check its output.
Do not trust it with things you know nothing about!
The informal poll results are in, & as I feared, most people are running ANOVAs in #R with functions that can give highly misleading (or flat out wrong) results. Follow along to see why using anova() and aov() is usually inadvisable & why you should be using car::Anova() 🧵