Tuesday, September 8, 2009

How does statistics hide facts?

Statistics often make us believe wrong conclusion and misled us into believing silly correlations.
Let me tell you a story that happened in Country Blue Chips (CBC).

A freshly minted assistant professor (in statistics) “discovered” that drinking orange+apple juice makes you smarter. He runs a regression analysis after a grant by the largest manufacturer of juices in CBC. He then engages 10,000 pupils from CBC primary school, and split them into groups of 10 comprising of 1,000 pupils per group.

Each group is given different compositions of orange and apple juice. E.g., group 1 drinks 10% orange juice+90% apple juice; group 2 drinks 15% orange juice+ 85% apple juice and so on. If the professor can have a large enough sample, he can have an unlimited number of sample groups. He then realized that hey, drinking orange juice and apple juice daily for 1 year does not increase the entrance score of the local gifted programme for 9 groups of pupils. However, coincidentally, the last group that drank 15.5% orange and 84.5% apple juice actually had a 99% percentage of people entering the gifted stream. “Okay, I am going to use this statistic and write a paper on it! Based on a 99% confidence interval, the hypothesis that orange+apple juice can make your child smarter is true! 99% went into gifted stream after drinking the juice composition!”

What about the rest that showed otherwise?!!

They are coincidences! Untrue!! I will not publish that! Anyway, people in CBC are too stupid to discover that!!

Assistant Professor gets a hefty grant and promotion to associate professor, the manufacturer of apple orange juice gets a boost from sales. Correlation does not mean causation. But who cares? At least not the people in CBC!

The same trick has been going on for many years in CBC. It is a sure win-win situation for the researcher and grant sponsor. I can also run the theory that TV makes babies smarter using the same manner. The variable is the number of hours of TV. For instance, I can choose groups of babies that watch 1 hour, 1.1 hour, 1.11 hour all the way to 10 hours of TV and choose the group that is the smartest, or with the highest percentage that goes into the gifted programme eventually.

Then I say, TV is good for your baby! Research has shown that 95% of babies who watched 2.145 hours of TV per day grow up to become scientists in CBC. (The group of babies who watched 2.144 hours of TV grew up to be ordinary Joe findings was not published)

Take a pinch of salt when you read statistics next time. Unfortunately, I cannot run a regression that tells me the amount of truth in all commercial statistics. But my rough guess?


At least in Country Blue Chips!


JW said...

Hi SgBlueChips

To share (I took from wiki):

"Lies, damned lies, and statistics" is part of a phrase attributed to the 19th Century British Prime Minister Benjamin Disraeli, among others, and later popularized in the United States by, among others, Mark Twain: "There are three kinds of lies: lies, damned lies, and statistics." The statement refers to the persuasive power of numbers, the use of statistics to bolster weak arguments, and the tendency of people to disparage statistics that do not support their positions.

PanzerGrenadier said...

Here’s wishing you a happy hari Raya Puasa holiday and a good long weekend catching up with loved ones, family and for some (sleep!).

Be well and prosper on your journey towards financial freedom.