more misuse of statistics

http://www.imdiversity.com/Villages/Woman/dialogue_opinion_letters/amoruso_alpha_girls_0313.asp
"ALPHA GIRLS: Is Today's 'Perfect' Seriously Flawed?"
by Carol Amoruso
retrieved December 31, 2008
IMDiversity.com

The article criticizes a new book (Alpha Girls by Dan Kindlon) which describes a new sort of girl, as it were: (about 20% of the 18-22 female cohort) young women who are confident, assertive - aggressive - driven overachievers (roughly). I have a complaint; let's start with some quotes.

"Surely, the picture is more nuanced than this. Nowhere is mentioned the fact—nor its impact--that 50% of first marriages end in divorce, 48% of all heads of households are unmarried, nor that 43% of today’s weddings (2002 figures, culled from U.S statistics by Divorce Magazine). These and other indications of great upheaval in the family dynamic cannot but redefine in dramatic ways a girl’s relationship to her father."

"Kindlon peers out from behind rose-colored glasses, while around him are stories of the frightening malaise gripping Generation Next’s young women. Overwhelmed by unreasonable expectations and emotional disconnect, they are characterized by Courtney E. Martin as “outwardly high-achieving and inwardly self-hating.” Writing for Alternet before the publication of her book, Perfect Girls, Starving Daughters, Martin reports that “7 million American girls and women have eating disorders, women account for twice as many panic disorders and depression and are plagued by 75% of autoimmune diseases.” Other sources cite that 75% of all anxiety attacks befall young women. Sadly, studies on growing narcissism among young women in particular predict impaired ability to achieve deep, intimate relationships."

There is something odd about this analysis. Amoruso is clearly aware that we're talking about a rather rare subset of young women - who are presumably rather different from the other 80%. And yet she is happy to take statistics that apply to young women generally, and assume that they apply just as much to this subset. Now, given the emphasis on good relationships with fathers discussed by Kindlon, wouldn't it be reasonable to ask: maybe these girls come disproportionately from non-divorced parents. Maybe the difference is that when these girls' parents got divorced, the fathers didn't step out of their lives entirely. In other words, this is an incomplete use of statistics: it leaves us with a lot of unanswered questions. The point was that these girls are unusual; the question ought to be: why?

the use and abuse of science

http://economix.blogs.nytimes.com/2008/12/05/why-does-us-health-care-cost-so-much-part-iii-an-aging-population-isnt-the-reason

"Why Does U.S. Health Care Cost So Much?"
by Uwe E. Reinhardt
December 5, 2008
The New York Times


This article is very nice, for some reasons. The one I'm most interested in is its use of these two graphs:



Basically, it argues that, by means of the first graph (combined with another which I omit), it can be made to appear that health care costs for the eldery are going to explode and add a ton of cost to the U.S. health care system. However, if we draw the graph differently, it becomes clear that this is not the case, and that the aging of the U.S. (as is stated) is a very gradual process.

The funny thing is that, if one thinks about it, even an 8% shift in the balance of elderly / non-elderly could have a tremendous impact on health care costs. That is, neither of these graphs really tells us anything. As it happens, the author backs up his claims with some studies that (purport) to show that increasing health care costs are not driven by aging. But the general point remains: it is very easy to lie with science, if one's audience is not attentive.

So, consider another graph provided by the author to prove his point:

This graph makes it appear that there is no correlation between the age-ratio of a population, and health care costs. But one might want to ask: what about controlling for other variables? Maybe if we control for, say, total population size, or economic output per capita, or who knows what, all of a sudden that cloud of dots will snap into a sharp line. So again, while it is a powerful image, there needs to be a lot more science and statistics done to back it up.

------------------------------------------

In other words, to return to some comments made earlier in this blog about the power of science: the lesson to be drawn is that it's easy enough to use the trappings of science - graphs, charts, statistics - to give the appearance of truth and validity to what are, in reality, questionable conclusions. But where do we go from here? Do we claim that science (as everything?) must rely on a degree of mutual trust - trust that people are working honestly and to the best of their abilities? Or do we say that people need to be more aware in their own dealings, and need to spend more time on analyzing the various facts they are presented with? (Do we all need to have taken courses on statistics?)