# @stem_antics on Instagram

- **Type:** Video
- **Original URL:** https://www.instagram.com/p/DdrOLcnRsTt
- **Gondola URL:** https://gondola.cc/posts/70844137-stem-antics-instagram
- **Thumbnail:** https://img.gondola.cc/tr:w-,h-,fo-auto/postThumbnails/564da6c3c4.jpg
- **Posted:** 2026-09-24T15:38:28.000+00:00
- **Account Owner:** Stem Antics (@stem_antics) — https://gondola.cc/stem_antics

## Caption

Correlation does not imply causation. Everyone has heard that one. The less famous twist is that causation does not imply correlation either.

A variable can directly drive an outcome and still produce a correlation coefficient near zero. The math and the mechanisms explain why.

- Nonlinear relationships
Pearson’s r only measures linear association. If y = x^2 and x is sampled symmetrically around zero, r comes out essentially zero. The dependence is perfect, yet the correlation vanishes.

- Cancelling pathways
Intense exercise can raise blood glucose through stress hormone release while also lowering it through increased glucose uptake by working muscle. Two causal paths with opposite signs can net to zero in the data. Statisticians call this a violation of faithfulness.

- Control systems
A thermostat is the classic example. The furnace clearly causes the house to warm, yet in a well-regulated home, furnace activity and indoor temperature look nearly uncorrelated, because the controller is actively erasing the effect it creates. Milton Friedman used this exact analogy to argue that effective central bank policy can look useless in the data.

- Restricted range
If every subject in a study receives nearly the same dose, the causal effect has no variance to appear in. Correlation needs variation to detect anything.

- Heterogeneous effects
A treatment that helps half a population and harms the other half can produce an average correlation of zero while strongly affecting every single individual.

The takeaway: correlation is a summary statistic, not a detector of mechanism. The absence of correlation is not evidence of the absence of causation. Real causal inference requires experiments, interventions, or explicit causal models, not just scatter plots.

#stemantics #statistics #causalinference #datascience #scienceeducation

## Stats

- **Views:** 4,173
- **Likes:** 32
- **Shares:** 0
- **Comments:** 1

## Tags

statistics, scienceeducation, stemantics, causalinference, datascience

---
Copyright (c) Gondola