The Grand Illusion
It’s not often that a F&SF writer can use one of his own books to show the shortcomings of statistics. However, as I write this, Solar Express has an “average” rating of three stars [3.2 stars, to be more exact], yet precisely one reader has given it a three star rating. Fifty-nine percent of the reader reviewers like it fairly well or a lot, and thirty-eight percent dislike it a little or a lot. So much for averages.
Yet as a society we tend to rely on statistics, all too often without really understanding what they mean. How often have you read a news item that states that eating something or using a certain product will increase the likelihood of getting cancer, or diabetes, or something else horrendous by ten or twenty or even a hundred percent? Yet do these statements ever point out the baseline risk?
For example, some advocates of using statin drugs [such as Lipitor] claim that use of statins reduces the incidence of heart disease by 50%. According to clinical studies over any five year period, roughly 2% of American males in the 50-60 age group will suffer a non-fatal myocardial infarction. Studies also show that statin use will reduce that rate to one percent. That is indeed a fifty percent reduction rate, but it’s only an actual risk reduction of one percent. Other studies showed that the decrease in mortality from fatal heart-related factors was offset almost completely in patients older than 70 by a corresponding increase in cancer deaths. But unless you or your doctor read the fine print in the studies, all you’re likely to hear is the fifty percent reduction in heart events. And if cancer runs in your family… well, you just might be better off not jumping at that “50% reduction.”
And take family income. In 2014 average [mean] family income was $72,641, but the median income [the amount where half the families make more and half make less] was only $51,939, or $20,702 – 40% less than the average. As a result, actually, about 67% of U.S. families make less than the “average.” Nor do such averages consider that one third of all American families live “paycheck to paycheck” and that 66% of those families are middle class with a median income of $41,000, well below the “average” family income.
Or take firearms. While there are 88 guns for every hundred Americans, all those firearms are actually in the hands of 43% of U.S. households.
Or… if you look at the Amazon stars, Solar Express is just an average book, despite the fact that only one person thought so.
As Mark Twain once famously observed, “There are three kinds of lies: Lies, damned lies, and statistics.”
I enjoyed “Solar Express” at a 5 star rating. One reason some may not enjoy it as much is because the philosophy and action so accurately reflect our current local, state. national and world politics. While the “gunfire” was perhaps not up to a comic book standard, it was definitely a very good thriller!
We have summed it up!
Well on goodreads is a little better:
https://www.goodreads.com/book/show/23848420-solar-express
(or worse? there are more 3 star ratings)
I quite liked the book, but it was lend to me by a friend, so this can be consider a lost sale? I will try to recommend but sadly not so many of my friends share the passion for Hard science fiction.
Actually on Goodreads (which imho is a considerably better indicator than Amazon, though it is owned by them) Solar Express has currently a 3.5 rating on 159 ratings with 30% at 3 and 32% at 4 so the distribution follows the rating
Similarly the family income distribution just shows that income is skewed towards the upper levels – though here the caveat is that these numbers are very rough insofar “real income” (including subsidies from tax deductions/rates to medical insurance to 401k’s, 529’s and the like as well as cost of living – especially housing) are not really taken into account
Statistics are a tool and it is actually a very useful one when applied properly, but it quickly degenerates into nonsense (even in very high level science as recent attempts to boost String Theory based on Bayesian analysis show) or politics when used improperly
And Solar Express was a superb book but I easily see how people expecting something else could dislike it
If people would understand3 things about statistics:
1. the difference between mean, median, and mode.
2. the difference between relative and absolute risk
3. how chi-squared worked.
then advertisers and politicians would have a much harder time lying to us with the truth.
@Peter Swartz: I saw what you did there. The groan you heard from the direction of northern california was mine. Well done.