Thomas Lumley 10/20/2013

Barren proxies

Read Original

The article discusses the statistical concept of 'barren proxies' in causal inference, where a measured variable is affected by a confounder but doesn't itself affect the outcome. It argues that in fields like medicine, most measured variables are technically barren proxies, but the key practical concern is whether the measurement reliably and closely tracks the true confounder under intervention, not its theoretical barrenness.

Barren proxies

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser