Pet peeve time.
On Thursday an analysis was released by the Mayo Clinic Proceedings based on data collected by the Aerobics Center Longitudinal Study regarding health effects of drinking coffee. To the credit of the data collectors, this study looked at 40,000 people over 16 years including ages ranging from 20 to 87 years old. Now you may have seen this in your newspapers, morning talk shows, and evening news with the headline “Drinking more than 4 cups of coffee increases mortality rate by 21%.”
Now I am fairly skeptical of these types of headlines, and chain emails too for that matter, which is why Factcheck.org and Snopes.com are web page favorites. The problem I have with the media drawing conclusions about data and research is obvious: they need to sensationalize the story, and nothing sensationalizes like fear (greed too). Also, I am not a statistician, but I know enough about statistical analysis to never take a news reader’s (yes that’s what they are) word on anything. I also, challenge polling and surveys as they are often suffer from study bias. Go ahead and read the questions in certain surveys and you will see how they can be constructed to generate the response the surveyor is HOPING to get.
But what about this study specifically gets my math geek nerd blood boiling? No causation. You see to make a direct link between coffee consumption and mortality rate, there needs to be statistically significant data to support that causation: drinking 4 cups of coffee causes higher mortality rate. What the study may have proven is correlation. Causation is pretty straightforward “Do ‘X’ and it will lead to ‘Y’”. Correlation on the other hand states “When ‘X’ goes up ‘Y’ goes up or down.” The difference is under correlation, there is no connection in tying ‘X” as the cause of ‘Y’.
Fact: There are more accidental swimming pool drownings in the summer.
Fact: More ice cream is consumed during the summer.
Conclusion: Eating ice cream can increase the odds of drowning.
Clearly that conclusion is erroneous, there is no causation, but there is correlation. I realize the example is outlandish, but it is to make the point that drawing causality conclusions from data can often be misleading especially in the hands of a newsreader.
Are we drinking too much coffee? I wouldn’t be surprised if we are as we seem to be consuming too much of EVERYTHING…right Mayor Bloomberg?
Perhaps after all, four out of five dentists don’t recommend Trident Sugarless Gum.