I’m at the JMP Discovery Conference this week, and one of the keynote speakers has me intrigued. Kaiser Fung, author of one of my RSS-fed favorites “Junk Charts“, compared learning statistics to learning about automobiles. He argued that most people would benefit from a working knowledge of how to interpret and use statistical inferences, and that they don’t really need to know how those inferences are derived.
“You don’t need to know how to build a car in order to know how to drive one,” he said. And he suggested that traditional statistics courses are akin to teaching students how to build a car. They’re focused on formulas, theory, and computation which most of us forget within a few weeks of the final exam. What we really need, he says, is an intuitive sense of how to use that statistical knowledge in our work.
Fung talked about how the public rejected new research recommending that women receive fewer mammograms due to an increased risk of cancer from the actual screening. The science is sound, but we choose not to believe it. Or we assume drug testing in major league baseball has a high probability for false positives, when in fact, many players who dope are tested repeatedly with negative results. Our intuition about statistics and probability are not only innocently wrong, in some cases, they’re willfully wrong.
But won’t we be better off if we at least learn how to understand statistical data? I may not be able to calculate the probability of a particular stock falling compared to the rest of the market, but I should have a sense of how that might affect my business and a plan to deal with it. As a consumer, I can’t define the sensitivity of a blood test to screen for cancer, but I do want to know how often I should expect a false positive.
So, a couple of questions:
- Have you taken any statistics courses? Did you learn anything?
- Do you regularly use what you’ve learned?
- What do you wish you understood about data, probability, or other statistical topics?