Daniel Oberski
daob.nl
Daniel Oberski
@daob.nl
Data science prof @ Utrecht. Incoming scientific director, Dutch national infrastructure for social sciences ODISSEI

Latent variables, Structural Equation Models, Measurement error

https://daob.nl | https://hds.sites.uu.nl | https://odissei-data.nl
Eurobarometer 38.1, 55.2, and 63.1. Provided without further comment.
June 25, 2024 at 9:28 AM
@michelenuijten.bsky.social I assume you're already looking into this, but just to let you know, I tried out ChatGPT and it coped perfectly well with the example on your website. Might be a bit worse at certain things, bit better on detecting non-standard ways of reporting. Prompt below
June 4, 2024 at 1:05 PM
(spoiler: the "secret" is to tune only eta)

The clickbaity intro obscures an important point: hyperparameters are not equally "tunable" (same goes for algorithms).

This was addressed in Probst, @boulesteixlaure.bsky.social's excellent article www.jmlr.org/papers/volum... (Fig 3)

Can recommend!
January 16, 2024 at 1:10 PM