Saturday, December 4, 2010

Say "hello" to the new Stat Girl on the Block

Media, TV, newspapers, but also online communities and blogs shape our opinions on health topics (all topics, really). In this busy world, we rarely take the time to read much beyond the sensational headline and the first couple of paragraphs. This is a dangerous trend, news items rarely contain the full information, journalists often cherry pick information from an original source and "fit" it to the desired headline (as eloquently described by Ben Goldacre here). Caveats and disclaimers are most often hidden at the bottom of the piece (Ben Goldacre again). References to original sources/publications, if they are given, are tedious to follow up. The whole exercise follows a stereotypical goal oriented (and the goal is NOT information) pattern (wonderfully spoofed by Martin Robbins here).

Evidence based bloggers, some online media and universities are pushing for less sensationalistic and more factual reporting. Recently, a campaign to add a "report an error" button to each news page was launched. Britain's National Health Service even runs a "Behind the Headlines" page that explains what is really in that paper.

Basically, as a "modern consumer" you need to know the basics of statistics, including the most important:

Correlation does NOT imply causality (or causal relationship).

We would probably suspect the (fictional) media headline "Storks deliver babies", based on the (true) fact that number of storks in Europe is strongly correlated with the local birth rate (the p-value is an impressive 0.008) wasn't true. But what about the many other health headlines, that tell us which herb/behaviour cures which disease, what is best for baby, what is best avoided for baby? Vaccines cause autism?

Enter Stat Girl, who "skewers the news, reminding the world that correlation is not causality one headline at a time".

With this "agenda" in mind, she is tackling the issues and outlets that target young mothers with their headlines.

Visit her blog and say "hello"! We need more analyses of media content and pressure for better reporting.

1 comment: