December 20, 2003
Smart Guy

Had lunch with friend Al yesterday, a smart guy who I'd called up because I wanted to have him explain some things about the threat of bioterrorism versus the threat of ordinary infectious diseases (an area in which he has some noteworthy expertise).

But Al's one of those peripatetic intellects I so enjoy, so we didn't get to bio stuff right away. Before we met, he emailed me some graphs of data he's been collecting on how much energy he and his wife consume at their house, with statistical comparisons to temperature, history over time, etc. And before I know it, Al's got his laptop out and he's explaining to me the statistical tests he uses and the different variables considered and the risks of data mining versus using statsitics to test the hypothesis with which you started.

Statistics has always been a weakness of mine, and Al loves to teach. Al pointed to one of the graphs, a line graph of month-by-month energy consumption over time for the average Albuquerque household, and pointed out that you can kinda see with your eye the generally upward trend amid the monthly ups and downs. I told him a favorite story of mine, about a smart scientist of my acquaintence who I'd asked to look at some climate data I was using for a story. The first thing the scientist did was hold up a printout of the graph at arms length and sort of squint at it. The eye, he explained, can perform some very useful first-order statistical tests.

Al told the story of the typical blood test your GP gives you. Normal for blood work is defined as 95 percent confidence interval, meaning in a random distribution of results, just by chance one in 20 will come back "abnormal". So if the doc's typical blood work on you measure 40 variables, just by chance in a normal person two will likely come back "abnormal". This is why it's important for at least your doctor to understand statistics, if not you. Better you.

Then I got onto the story I'm working on about the diurnal temperature variation and the signature of global warming, and Al in his usual penetrating way asked me a series of questions that drilled down right to the heart of the issue.

And when we finally got to bioterror and infectious diseases, it was almost an afterthought. I was ready for a bit of a donnybrook over my theme - that infectious diseases are far more threatening than bioterror, and we're therefore spending our energy on the wrong thing. Al, who I expected to disagree, granted my point up front, wtih the caveat that money spent on the bioterror threat yields equivalent gains on the infectious disease front, so it's a wash in terms of the overall research effort, and a net benefit if the bioterror threat means more total money is spent.

Posted by John Fleck at December 20, 2003 08:31 AM
Comments

Your recounting of the story about the GP reminded me of this entry that Jacques Distler made a while ago (back in September):

http://golem.ph.utexas.edu/~distler/blog/archives/000229.html

(ok, that is a link to story containing a link to *the* story, but I want to keep my sources straight). It appears that, despite the logical hope that doctors should understand statistics for the purposes of interpreting probabalistic test results, this is not always a valid assumption.

Posted by: Malcolm Tredinnick on December 22, 2003 07:20 PM

OK, I got the first one right, but my brain hurt too much and I admit I just bagged the second. But I'm not an MD. As my friend Al puts it, "MD stands for mathematically deficient". And Al's an MD.

Posted by: John Fleck on December 22, 2003 08:57 PM
Comments
Comments