Something to think about as you build your equations for statistical analysis:
"If you control for lung cancer, smoking doesn't kill you."
Via Layna Mosley who credits it to a discussant on one of her ISA panels.
IPE @ UNC
IPE@UNC is a group blog maintained by faculty and graduate students in the Department of Political Science at the University of North Carolina at Chapel Hill. The opinions expressed on these pages are our own, and have nothing to do with UNC.
Bookshelf
Tags
Blog Archive
-
▼
2010
(478)
-
▼
February
(45)
- The Non-End of Influence
- International Relations
- Politics isn't the Dispassionate Quest for Truth
- Hell Freezes Over
- Economics of the Somali Pirate Business Model
- Who Wants to Elect a President?
- Tragedy of the Day
- Update on 4% Inflation Target
- A Little Light Game Theory (Greek Sovereign Debt E...
- It's the Economy, Stupid. Or Is It Politics?
- Inconsistent Thoughts on Time-Inconsistency Problems
- US Navy and Pirates!
- African Coups and Colonels
- Quote of the Day
- Envy, Altruism, and Trade Policy
- Transparency and Missing Data
- One Last ISA Post
- Fact of the Day
- Welcome
- ISA
- Why do governments buy companies when they could j...
- New Stata Data Repository
- Delayed Live-Blogging of the IR and Blogging Panel
- Game Day
- Where Dat?
- Damn Greeks (or Why Germany Never Wanted EMU in th...
- The IMF Doesn't Like Inflation Any More than the F...
- ISA
- Don't Make Governance Political
- The New Cold War
- Quote of the Day
- Vreeland on the "Bad Ass" Theory of Dictators and ...
- Quote of the Day
- Fly the Flag
- ECB on Wire
- Coulda Seen This One Coming
- Is Greece Too Big to Fail?
- PIRATES!
- It Depends on Your Posterior Distribution
- Predicting History
- No Way Out?
- What Does Unsustainable Mean?
- There Is No Resource Curse
- Courage
- Ice, Ice, Baby (or a Lesson in Bureaucracy)
-
▼
February
(45)
Tuesday, February 23, 2010
Subscribe to:
Post Comments (Atom)
2 comments:
Sweet!
Although, I'm not sure how that really works. Was the person saying that multicollinearity would make smoking insignificant if both variables are on the right-hand side? Not necessarily, b/c smoking causes emphysema and heart disease as well. In fact, the dangers of smoking are mostly *not* lung cancer, so controlling for it might be perfectly appropriate.
The quote might not be entirely accurate, but the idea behind it is important: sometimes controls can obscure rather than enlighten.
Post a Comment