OK, this is an oldie but goodie from xkcd, but I saw it in a recent presentation and couldn’t resist sharing.
Of course, you’d never know that from the sensationlist press.
If you enjoyed this post, make sure you subscribe to my RSS feed!
Twitter Feed- RT @nicolastorzec: Introducing the LinkedIn Data Science team, by @dtunkelang: http://t.co/hyyzYT7u - Nice initiative & good PR May 21, 2012 alter_lego (Luís Pinheiro)
- RT @uwengineering: Tx! RT @dtunkelang: Wow, UW is on a tear! Congrats on hiring Jeff Heer, Daniela Rosner, Carlos Guestrin, and Emily Fox! http://t.co/El4nO0gn May 21, 2012 landay (James Landay)
- Introducing the LinkedIn Data Science team, by @dtunkelang: http://t.co/hyyzYT7u - Nice initiative & good PR May 21, 2012 nicolastorzec (Nicolas Torzec)
- Happy 2c my Uni classmate Darya Sorokina among you guys RT @dtunkelang Check out the awesome data scientists on my team http://t.co/c7WXCFyp May 21, 2012 mariagrineva (Maria Grineva)
- Tx! RT @dtunkelang: Wow, UW is on a tear! Congrats on hiring Jeff Heer, Daniela Rosner, Carlos Guestrin, and Emily Fox! http://t.co/El4nO0gn May 21, 2012 uwengineering (UW Engineering)
- RT @mrogati: From data product innovation to writing books on R: @dtunkelang highlights the data science team he built at LinkedIn: http://t.co/p4zjaCLL May 21, 2012 dtunkelang (Daniel Tunkelang)
- From data product innovation to writing books on R: @dtunkelang highlights the data science team he built at LinkedIn: http://t.co/p4zjaCLL May 21, 2012 mrogati (Monica Rogati)
- Wow, UW is on a tear! Congrats on hiring Jeff Heer, Daniela Rosner, Carlos Guestrin, and Emily Fox! http://t.co/HdjlLm9u May 21, 2012 dtunkelang (Daniel Tunkelang)
- The Noisy Channel http://t.co/Mfi2sZWu May 21, 2012 asyaasia (asia)
- #DataScience at #LinkedIn -meet the Linked In Data Science team http://t.co/Ucv4IfJ8 May 21, 2012 KDRrecruitment (KDR Recruitment Ltd)
6 responses so far ↓
1 Milan // Nov 17, 2008 at 5:33 pm
Of course, since that was first published there’s been more than a bit of self-referential measurement error introduced. The current count for “skydiving” results is 633, for “blogging” 652, and for “knitting” more than 10,700.
Whether that indicates high overlap in the sample sets or a deadly wave of recent yarn & needle fatalities, I can not imagine.
2 Daniel Lemire // Nov 17, 2008 at 9:24 pm
Blogging leads to weight loss? Who knew?
Where do these “journalists” take their info? I know plenty of overweight bloggers.
3 Daniel Tunkelang // Nov 17, 2008 at 10:44 pm
I can’t speak to the rigor of journalists–or bloggers–in obtaining their statistics.
But, in all seriousness, the techniques that some people have discussed for learning from distributional similarity in documents and query logs are strikingly in line with that xkcd comic.
It does make you (or at least me) wonder how to game the log-mining approaches by spamming the query log, much as spammers already exploit distributional similarity to create documents that get past spam filters.
4 jeremy // Nov 18, 2008 at 7:45 pm
Be careful about what assumptions you draw based on Google hit and/or suggestion counts.
There is ongoing evidence that Google does not report the real numbers.
http://sethf.com/infothought/blog/archives/001402.html
5 jeremy // Nov 18, 2008 at 7:46 pm
Here is a better link:
http://www.hyperorg.com/blogger/2008/11/12/obama-v-bush-google-counts/
6 Daniel Tunkelang // Nov 19, 2008 at 1:15 pm
Seth Finkelstein–haven’t seen him since I was an undergrad at MIT! But I do agree with him and Dave Weinberger that Google’s counts are systematically skewed, though it’s not clear what causes that skew. Mr. Schmidt, tear down this black box!
Leave a Comment