It happened in a seminar yesterday at my institute that somebody bullied a phd student to change an analysis, because two of the predictors were correlated. Multicollinearity is a brain worm, isn't it? It never makes sense to choose control variables based on pre-model correlations.

Show Post

The problem is noise, not N: why increasing the sample size isn’t all that. statmodeling.stat.columbia.edu/2026/10/10/t...

Show Post

In a typical meeting room, it’s wild to see how quickly CO2 levels get into the range where folks are no probably no longer thinking clearly. I don’t understand why as a society we haven’t bothered to install air quality monitors in workplaces and other public spaces.

Show Post

link.springer.com/article/10.3...

Show Post

Much appreciation to @matti.vuorre.com @rmkubinec.bsky.social @chelseaparlett.bsky.social for helping craft this beast! Also to reviewers @scoretta.bsky.social and @solomonkurz.bsky.social for making this a better paper!

Show Post