Tags: Education | Health Topics | Mexico | correlation

Correlations Can Be Fun and Profitable, or Neither

Image: Correlations Can Be Fun and Profitable, or Neither

(Alain Lacroix/Dreamstime)

By
Friday, 02 Sep 2016 03:02 PM Current | Bio | Archive

One of the first things they teach you in a statistics class is: “Correlation between two sets of numbers suggests but does not imply causation.”

There are many websites out there that entertain readers by charting out outrageous, spurious correlations. For example, Harvard law student Tyler Vigen lists a smattering of amusing correlations at twentywowords.com.

Some favorites: Increasing ice cream consumption tracks an equally steady rise in the murder rate. But don’t worry, a decline in the market share of the Internet Explorer browser parallels a decline in U.S. murders. Also, importing more Mexican lemons seemingly prevents highway deaths.

Some correlations, as ridiculous as they are, have lasted for decades. Take New York City’s “Pizza Principle,” which maintains that there’s a direct correlation between the rising cost of a plain slice of pizza in New York and a single public transit fare.

When it comes to correlating environmental factors (particularly chemicals) with diseases, the sky’s the limit. The neurodegenerative condition known as Parkinson’s disease is a favorite for researchers looking for correlations.

Following the influence epidemic of 1916-1917, there was an increase in a neurologic syndrome that included Parkinsonism, so an infectious agent was suspected. None has ever been found. Then there was a correlation found between the early onset of Parkinson’s disease with people who drank well water in childhood.

A 1992 article in the journal Food and Chemical Toxicology noted a correlation between Parkinson’s disease and working in a paper mill or in an orchard.

Perhaps a similar chemical could be found in both environments.

But no causative agent has in fact been discovered, even though there are such related tantalizing correlations as those unearthed by one researcher named Tanner, for example, who found that people who developed Parkinsonism before the age of 48 were more likely to have lived in a rural town with a population of less than 1,000.

Sometimes strong correlations lead to a conclusion too disturbing for some researchers.

Take as an example French sociologist, social psychologist and philosopher Emile Durkheim. Every freshman college student in a Sociology 101 class has Durkheim and his alleged masterwork, “Suicide” (1897) foisted upon them. Durkheim thought that Protestants exhibit higher rates of suicide than do Catholics, and that uneducated people were subject to higher suicide rates than educated folk.

However, a careful 1983 analysis by Rodney Stark, Daniel P. Doyle, and Jesse Lynn Rushing in the Journal for the Scientific Study of Religion puts forward the idea that Durkheim fudged his data to allow a better fit with his preexisting ideas. Clearly the value of correlations can depend on the prejudices of the supposedly disinterested researcher wielding the statistical tools to “uncover” them.

That's not to say that some correlations are not just amusing, but intriguing. Decades ago, the U.S. military conducted an experiment where a cafeteria was fitted out with a series of vending machines stocking every conceivable flavor of soft drink. People who were allowed in the cafeteria had extensive dossiers as to their physical and mental condition, and a video camera observed what they purchased from the vending machines.

The result? People who had been diagnosed with neuroses or other mental abnormalities had a tendency to buy... grape soda!

Perhaps the biggest area of study involving correlations is the stock market. Everyone is looking for a magic key to unlock a lucrative investment strategy. Various software trading platforms have features that attempt to uncover correlations among traded assets. One can get lost in this world, and discover all sorts of amusing correlations that are as spurious as the ones described at the beginning of this blog posting.

For example: Business Insider used once such platform (Select Sector SPDRs' Correlation Tracker) to note the fact that General Motors’ stock tracks that of PowerShares PIO index, which tracks companies that make products designed to conserve and purify water for homes, businesses and industries.

An interesting correlation. Is there an underlying cause related to both? Probably not.

One could say the same for the remarkably good correlation between Disney and the U.S. Industrial Sector.

Amateur investors have lost fortunes both large and small by discovering and betting their life savings on such correlations.

One day back around 1980, a researcher came into the student library at Rutgers University on the Newark, New Jersey campus where an acquaintance of mine worked. The gentleman had spent his life researching correlations and cyclic behavior of everything and anything, presumably to give him an edge in his investments.

My friend asked him, “What is the strangest correlation you have ever found?”

To my best recollection, the answer went something like this: As part of state mosquito control programs throughout U.S., the appearance and rate of mosquito larvae hatching is measured in breeding areas such as slow streams, shallow lakes, reservoir margins, and so forth.

This gentleman said that he had hit upon a pretty good correlation showing that the number of hatching larvae decreased whenever there was a nuclear weapons test. (The 1963 Limited Test Ban Treaty stopped nuclear testing in the atmosphere, underwater or in outer space, but France continued atmospheric testing until 1974, China until 1980, and America kept up underground testing until 1992.)

In any case, my friend said, “So, the nuclear tests somehow reduce the number of mosquito larvae hatching?”

The man replied, “No, actually the number of larvae hatching goes down two or three days before the nuclear test.”

Putting aside the possibility of precognitive mosquito larvae, as I have said, if you engage in this kind of activity too long, it’s easy to get mentally lost in the world of ideas.

Richard Grigonis is an internationally known technology editor and writer. He was executive editor of Technology Management Corporation’s IP Communications Group of magazines from 2006 to 2009. The author of five books on computers and telecom, including the highly influential "Computer Telephony Encyclopedia" (2000), he was founding editor-in-chief of Jeff Pulver’s Voice on the Net (VON) magazine from 2003 to 2006, and the chief technical editor of Harry Newton’s Computer Telephony magazine from its first year of operation in 1994 until 2003. Read more reports from Richard Grigonis — Click Here Now.

 


 

© 2017 Tribune

   
1Like our page
2Share
RichardGrigonis
One of the first things they teach you in a statistics class is: “Correlation between two sets of numbers suggests but does not imply causation.”
correlation
1028
2016-02-02
Friday, 02 Sep 2016 03:02 PM
Newsmax Inc.
 

Newsmax, Moneynews, Newsmax Health, and Independent. American. are registered trademarks of Newsmax Media, Inc. Newsmax TV, and Newsmax World are trademarks of Newsmax Media, Inc.

NEWSMAX.COM
America's News Page
© Newsmax Media, Inc.
All Rights Reserved