Correlations in databases

Thursday 30 December 2010

Cause & effect

We get frequent news items that claim something is related to another factor. 

Students need to learn how to read through these messages for better understanding of the science and consequences.

Dave Mosher writes for Wired Science and produced an article [1] and has produced this list of correlations that may be statistically okay but not have any real meaning and may not demonstrate a true cause and effect relationship.

According to the article, finger length is related to two genes that also control the growth of male testes in utero.  Rosalind Eeles and her team at the Institute for Cancer Research (ICR) studied prostate cancer incidences in over 1,500 men  versus 3,000 random men.

"Men older than 60 years with a longer index finger were 33 percent less likely (on average) to develop prostate cancer. Younger men with a longer index finger fared even better, with an 87 percent average reduction in risk." (Mosher, 2010)

As students learn in topic 1- Statistics, a correlation does not necessarily mean there is a direct cause, so studies such as this needs to be repeated very carefully to truly establish the relationship but for now medical statistics may lead doctors to do screenings on those more at risk.


Footnotes

  • 1. Mosher, David. "The Weirdest Indicators of Serious Medical Risks | Wired Science | Wired.com." Wired.com . N.p., n.d. Web. 30 Dec. 2010. .

Tags: correlation, cause, effect, cancer,