Misinformation: Statistics

Economist and journalist Sanne Blauw's TEDx talk on how to watch out for misleading statistics reported in the news.

While the presence of numbers and statistics in a news story may make the evidence look straightforward, it's important to be thoughtful about interpreting them.

• How carefully were the data collected?
• Is it measuring what the writers claim it's measuring?
• What are the implications of the results?

Let's look at an example from the "Carrying Some Extra Pounds May Not Be Good After All". From the NPR article:

[Boston University School of Public Health demographer Andrew Stokes'] group found a 6 percent increased risk of dying from any cause among individuals with a history of being overweight.

Although Stokes says that 6 percent "is only a modest increase," it's still "extremely worrisome" because so many Americans are overweight.

It can take a lot of research to definitively track down the cause of something. Two variables might look related, but it's important not to jump to conclusions.

How are ice cream sales and shark attacks related? Does eating ice cream make you more delicious for sharks to attack? That's not the real relationship here.

There's a correlation between ice cream sales and shark attacks, but one is not causing the other.  So why do their graphs look similar?  In hotter weather, more people are eating ice cream and spending time at the beach, where sharks can attack them.

For real shark attack statistics, see the Florida Museum's International Shark Attack File. Shark attacks aren't really as likely as this made-up graph suggests. For example, in 2000 beach attendance in the U.S. reached 118,000,000, but there were only 41 shark attacks.

For more examples, see the website Spurious Correlations, which uses real data to find correlations between unrelated variables: