Finally, a slow news day. Even the
Huffington Post's home page is running the same days-old stories, including Obama's 12-stitch, elbow-to-lip cut in a pickup basketball game with friends. And including what has now seemingly become obligatory: the 6 -- count them every day -- always 6 and sometimes even 7 or 8 -- blurb stories with pictures of She Who Shall Not Be Named. It's really tiresome, and she doesn't deserve that much coverage. We can only hope the excess will hasten the burn-out of public fascination.
Anyway, I've been saving this up for just such a slow-news day. One of my hobby-horse complaints is how poorly journalists understand statistics and how they can be manipulated to show almost anything you want. The public is often misled by well-meaning news reporters who simply don't understand the basics. And then there's the whole other realm of the partisan cherry-picking numbers game of junk science and political lying.
Now comes news of a book,
The Numbers Game:
The Commonsense Guide to Understanding Numbers in the News, in Politics, and in Life, by Brits Michael Blastland and Andrew Dilnot, who base it on a popular BBC radio show.
A
NYT review begins by debunking, and explaining, some headline grabbers:
"Most people have more than the average number of feet." How's that? Simple. Because of amputations and birth defects, some people have only one foot; and almost no one has more than two. Therefore, the average is somewhat less than two. Therefore, the vase majority of people, having two feet, actually do have more than the average. Get this picture?
Another:
"Republicans enjoy sex more than Democrats." Facts: more men than women vote Republican, and men tend to report enjoying sex more than women do. Therefore, Republicans are more likely than Democrats to report enjoying sex.
And a common one from the daily news:
"Poll shows majority favor X." Read further, and the numbers show 45% favor X, 43% favor Y, and 12% are undecided. A majority? No, that could at best be called a plurality; but even that is misleading, because the reliability will likely be something like + or - 4%. So it's really a statistical tie. Or sometimes, even correctly calling something a majority, can be very misleading: 51% to 49% is indeed a majority, but the implication of "most people favor" is anything but true. "Majority" is not equivalent to "Most People."
But then you get into reported results of actual or so-called "studies" with real "results" in numbers that can still be very misleading. A lot depends on how representative the sample, on how the questions are asked, and on the what and how of data collection. Some of my favorite examples stem from my long-running battles with anti-gay rhetoric and even in my battle against the pope and his position on condoms last spring.
One of the "studies" that religious conservatives love to quote comes from multiply discredited psychologist Paul Cameron, who claimed that gay men should not be allowed to adopt children because they have a shortened life span. What did he base this on? A "study" in which he recorded the age at death of noted men who had died from AIDS and were listed in the obituary column of a popular gay magazine. Then he compared the age at death of this highly selective sample of gay men with the national average of life expectancy of all men. Ergo, these men died at a younger age; ergo no gay man should be entrusted with responsibility for a child because he will die young.
What is obviously wrong with this? The sample. This is not a random sample of equal numbers of gay men and straight men and comparing the age at which they died. This is a highly skewed sample of gay men taken from a list of those who had died young
because of one specific cause and comparing it with the general population of men. The result that Cameron claims would have to assume that all gay men are infected with HIV and that all straight men are not.
If you actually did an honest, age-of-death comparison, there would be many factors besides HIV that would affect the results, some of which might favor gay men and some straight men. Such things as the incidence of engaging in high risk sports, violent crime, being soldiers in combat, tendency toward traffic accidents, suicide rates, as well as the incidence of drug-related HIV spread in straight men. Complicated? You betcha.
Another example: In response to my op-ed piece in the
AJC in which I was scathingly critical of the pope's statement that relying on condoms actually increases the spread of HIV, a letter from the head of Catholic League for Religious and Civil Rights of NY, stated as fact that "the promiscuous distribution of condoms has coincided with a precipitous increase in infections." He doesn't, of course,
claim causation; but he ends with: "The holy father can connect the dots; why can't Roughton?" It was clearly meant to imply causation, and the average reader will assume that is what is claimed.
The is the simplest form of mistaking coexistence with causation. It's true, back in the 80's when the rise of HIV infections was rampant and rising rapidly, there was a lag in any effective measure showing up in reduced new cases -- simply because there is often a long period between contracting the virus and diagnosis. This year's "new cases" may have become infected two years ago. However, at this same time there was a concerted effort to distribute free condoms on a wide spread basis to help stem the spread. So the number of "new cases" may actually have gone up during the first year or so of free condom distribution.
So, yes, the rising curve of delayed-reporting of new cases coincided with the institution of free condom distribution. But, rather than the cause of the spread, it was a belated effort to stem the spread. And any positive effect of condom use may not show up in the new-case statistics for a year or two. And now, longer-range statistics, especially in Africa, have shown a direct, robust correlation of condom use and
reduction of the spread of HIV, just as I stated and as the pope and his minions contradicted. Public health authorities now confidently assert that condom use is the most effective single factor in reducing the spread of HIV.
I believe it was Harry Truman who coined the phrase: "Lies, Damned Lies, and Statistics."
This book should be interesting reading, even for those who don't need to be taught about the numbers game.
Ralph