🫗Confounder

A third variable that influences both the thing I think is the cause and the thing I think is the effect, so that the two move together without one driving the other.

From 🎓 Lecture 13 - Confounding and what regression cannot tell you. A regression will report the association with a straight face. It cannot tell me what I left out.

Origin: Latin confundere, to pour together, from con (together) and fundere (to pour). Two effects poured into one glass.

An example from my work, added 18 November. In 📌 Statistics II - Project - What predicts vessel delay at Greywater one of my predictors looked as if it explained delay. Hana Kobayashi read the draft and asked what time of year those rows came from.

The time of year was moving both: the predictor and the delays. Once it was in the model the neat effect shrank to something I could not distinguish from nothing. I had been warned in September and walked into it in November anyway.