One of the intellectually fun aspects of taking BIOL 607 is that the course has me considering statistics and probability everywhere. Often it is in the context of reading some news story and then questioning the design of the study or the validity of the conclusions being reported. I am struck by how poorly understood statistics and probability are for most people. If, as our textbook points out, even otherwise well-educated physicians often mistake the probabilities associated with many diagnostic tests, how well is the typical adult able to make sense of all the statistical “noise” emanating from drug companies, politicians or financial pundits? To put the challenge in perspective, according to the US Census Bureau, in 2009 49.7% of the US population had less than one year of college education and only 27.9% had attained a Bachelor’s degree or higher (with a margin of error of +/- 0.1 percent). No wonder companies, politicians and snake-oil salesmen can so easily throw up a bunch of numbers and cow the general public into believing they are valid or have meaning.
One of the pernicious aspects of statistics is how attaching numbers to an assertion can make anything seem “scientific” or “accurate” or “professional”. The response by most people is to be impressed and to believe whatever is being asserted without questioning the statistics or even understanding them. So much the better if you can add several decimal places to the numbers –clearly you must have measured and analyzed with great precision, right? But as Nate Silver is quick to point out in the first chapter on the 2008 financial meltdown, just because the numbers are calculated with great precision, doesn’t mean they are the least bit accurate. In the computer field one quickly learns GIGO – garbage in, garbage out. Computers have made it easy to crunch data and spit out spiffy-looking charts and plots and all manner of impressive statistics but the software has no way of knowing the quality or even the applicability of the data being entered. In its most benign form all these numbers are a form of puffery – making trivial or irrelevant data seem like something. At times statistical analyses can convince us we know more than we do or lead us to faulty conclusions. More dangerously, statistics can be deliberately used to obfuscate and outright lie.
It is such a different world that we in BIOL 607 inhabit, struggling as we may be with plotting results in R or simulating results of a thousand samples, but at least able to speak and understand the language of statistical and probabilistic reasoning (admittedly some, like this author, rather less fluently). It was a very statistically fluent audience on Tuesday at Boston University, where I sat listening to Dr. Andrew Pershing of GMRI discuss the statistical manipulations he had used to tease out the signal from the noise in order to model ocean warming in the Gulf of Maine. Well-executed statistical work was on display everywhere at the RARGOM Annual Science Meeting and thanks to BIOL 607 I found myself focused on the standard deviations, confidence intervals, and experimental designs like never before.
My work as a science educator leads me to always be on the lookout for potential teaching examples – now days I am hot on the trail of examples of statistical “malpractice”, poor experimental design or erroneous assumptions/conclusions in science reporting. These examples I will end up using with the teens and the teachers I teach in my programs. To this end, I welcome any examples from all of you and I’ll share a graph encountered this week:
From Easterbrook, Greg, “What Happens When We All Live to 100” in The Atlantic, October 2014, p.65.
So what’s wrong here? A number of things, not the least of which is that the average person will read either of the trend lines extending into the future as determinant – as facts – happily continuing upward toward ever longer life expectancy. As scientists we know that there is nothing to prevent that life expectancy trend line from curving decidedly downward in future years or at least flattening out, but the implication here is that our children and grandchildren are “guaranteed” to look forward to ever longer lives. But will they?
Thanks to Jarrett Byrnes, Nate Silver and Whitlock & Schluter, statistics – both good and bad – now haunt me everywhere…