This chapter began as two radio talks delivered on the ABC,
almost thirty years ago. My friend Peter Chubb asked me if I had addressed
these issues, and I said that I hadn’t, but that I had provided a link to the
text of the talk. Two nights later, I decided to add it, the next night, I
rewrote it.
There is enough information here to let readers try the following exercise in Evil
Statistics out for themselves.
*
Boris, Don and Tony
went fishing, and caught ten fish. Four weighed 1 kg, two were 2 kg, two were
three kg, one was 6 kg, and one was 10 kg. They reported that the average was 1
kg, 2 kg and 3 kg, and all were telling a sort of truth. Boris reported the mode, the most common mass, Don reported
the median, the mass of the middle
two fish, Tony reported the mean,
adding all the masses and dividing by ten. Each value was true, each was
different.
It all sounds a bit like “Lies, damned
lies, and statistics”, but who first said that? The popular myth is that it was Mr.
Disraeli, the wellknown politician, but many quite reputable and
reliable reference books blame author Mark Twain.
It turns out that it was first
published by Twain all right, but Twain attributed the line to Disraeli,
and you won't find the story in any earlier publication than Twain's autobiography.
In short, Mark Twain made the whole thing up! Disraeli never spoke those
words: Twain invented them all, but he wanted the joke to have a greater force,
and so gave the credit to an English politician.
Twain wasn't only wellknown for his
admiration of a good “Stretcher” (of the truth, that is), he even lied when he
was talking about lies, and his name wasn't even Mark Twain, but Samuel
Clemens! Now would you buy a used statistic from this man?
Last century, when Disraeli is supposed
to have made the remark, statistics were just numbers about the State. The
state of the State, all summed up in a few simple numbers, you might say.
Now governments being what they are, or
were, there was more than a slight tendency in the nineteenth century to twist
things just a little, to bend the figures a bit, to bump up the
birth rate, or smooth out the death rate, to fudge here, to massage
there, to adjust for the number you first thought of, to add a small
conjecture or maybe to slip in the odd hypothetical inference.
It was all too easy to tell a few
small extravagances about one's armaments capacity, or to spread the
occasional minor numerical inexactitude about whatever it was rival
nations wanted to know about, and people did just that. Even today, when
somebody speaks of average income, if you don’t smell fish, at least remember
them, and ask if that’s the mean, the median or the mode.
When I was young, I smoked cigarettes,
but the cost and the health risks convinced me, so I stopped, back in 1971.
Smokers think we reformed smokers are tiresome people who keep on at them,
trying to get them to stop as well.
The nonsmokers say those who still
puff smoke are the tiresome people, who can't see the carcinoma for the smoke
clouds, who deny any possibility of any link between smoking and anything. Like
the tobacco pushers, the smokers dismiss the figures contemptuously as “only
statistics”. The really tiresome smoker will even say a few unkind things about
the statisticians who are behind the figures. Or about the statisticians who lie
behind the figures.
By the end of the 19th century,
statistics were no longer the mere playthings of statesmen, they were way to clump
large groups of related facts into
convenient chunks. If you can see how the statistics were arrived at,
perhaps you can trust them.
At one stage in my career, I led a gang
of people who gathered statistics and messed about with numbers, but we preferred
to be called ‘numbercrunchers’. People say a statistician is “somebody
who's rather good around figures, but who lacks the personality to be an
accountant”.
They speak of the statistician who drowned in a lake with an
average depth of 15 cm. We are told that a statistician collects data and draws
confusions, or draws mathematically precise lines from an unwarranted
assumption to a foregone conclusion. They say “X uses statistics much as a
drunkard uses a lamppost: rather more for support than for illumination”.
Crusty old conservatives give us a bad name, pointing out
that tests reveal that half our nation's school leavers to be below average,
which is true, but it is equally true that the vast majority of Australians
have more than the average number of legs. All you need is one Australian
amputee!
If somebody does a Little Jack Horner with a pie that's
absolutely bristling with statistical items and they produce just one
statistical plum, I won't be impressed at all: the plum's rather more likely to
be a lemon, anyhow.
The statistics have to be plausible and significant. Later, I
will show you a statistical link between podiatrists and public telephones:
this is obviously nonsense, and we will ignore it. There is no logical reason
for either to influence the other.
Still, unless there is a plausible reason why X might
cause Y, it's all very interesting, and I'll keep a lookout, just in case a
plausible reason pops up later, but I won't rush to any conclusion. Not just
yet, I won't.
First, I will check on the likelihood of a chance link,
something we call statistical significance. After all, if somebody claims to be
able to tell butter from margarine, you wouldn't be too convinced by a single
successful demonstration, would you? Well, perhaps you might be convinced: certain advertising agencies think so, anyway.
If you tossed a coin five times, you wouldn't think it meant
much if you got three heads and two tails, unless you were using a
doubleheaded coin, maybe. If somebody guessed right three, or even four, times
out of five, on a fiftyfifty bet, you might still want more proof.
You should, you know, for there's a fair probability it was
still just a fluke, a higher probability than most people think. There's about
one chance in six of correctly guessing four out of five fiftyfifty events. Here
is a table showing the probabilities of getting zero to five correct from five
tosses:
zero right

one right

two right

three right

four right

five right

1/32

5/32

10/32

10/32

5/32

1/32

The clever reader may notice a
resemblance to Pascal’s triangle here!
Now back to the butter/margarine study. Getting one right out
of one is a fiftyfifty chance, while getting two right out of two is a twenty
five per cent chance, still a bit too easy, maybe. So you ought to say “No,
that's still not enough. I want to see you do it again!”.
Statistical tests work in much the same way. They keep on
asking for more proof until there's less than one chance in twenty of any
result being just a chance fluctuation. The thing to remember is this: if you
toss a coin often enough, sooner or later you'll get a run of five of a
kind.
As a group, scientists have agreed to be impressed by
anything rarer than a one in twenty chance, quite impressed by something
better than one in a hundred, and generally they're over the moon about
anything which gets up to the one in a thousand level. That's really strong
medicine when you get something that significant.
There. Did you spot the wool being pulled down over your
eyes, did you notice how the speed of the word deceives the eye, the ear, the
brain and various other senses? Did you feel the deceptive stiletto, slipping
between your ribs? We test statistics to see how “significant” they are, and
now, hey presto, I'm asserting that they really are significant. A bit
of semantic jiggerypokery, in fact.
And that's almost as bad as the sort of skullduggery people
get up to when they're badmouthing statistics. Even though something may be
statistically significant, that's a long way away from the thing really being scientifically
significant, or significant as a cause, or significant as anything else, for
that matter.
Statistics make good
servants but bad masters. We need to keep them in their places, but we oughtn't
to refuse to use statistics, for they can serve us well. Now you are ready to
object when I assert that all the podiatrists in New South Wales seem to be
turning into public telephone boxes in South Australia, and it all began with
Florence Nightingale. Most people think of her as the founder of modern
nursing, but as part of that she created ways to use statistics to pinpoint
facts.
After her name was made famous, directing nursing in the
Crimean war, she returned to London in 1857, and started to look at statistics,
and the way they were used. She wrote a pamphlet called “Mortality in the
British Army”, and the very next year, she was elected to the newly formed
Statistical Society.
She looked at deaths in hospitals, and demanded that they
keep their figures in the same way. The Statistical Congress of 1860 had, as
its principal topic, her scheme for uniform hospital statistics. It isn’t
enough to say Hospital X loses more patients than Hospital Y does, so
therefore Hospital X is doing the wrong thing.
We need to look at the patients
at the two hospitals, and make allowances for other possible causes. We have to
study the things, the variables, which change together. Statistics, remember,
are convenient ways of wrapping a large amount of information up into a small
volume. A sort of shorthand condensation of an unwieldy mess of bits and
pieces.
And one of the handiest of these shorthand describers is the
correlation coefficient, a measure of how two variables change at the same
time, the one with the other. Now here I'll have to get technical for a
moment. You can calculate a correlation coefficient for any two variables,
things like number of cigarettes smoked, and probability of getting cancer.
The correlation coefficient
is a simple number which can suggest how closely related two sets of
measurements really are. It works like this: if the variables match perfectly,
rising and falling in perfect step, the correlation coefficient comes in with a
value of one. But if there's a perfect mismatch, where the more you smoke, the
smaller your chance of surviving, then you get a value of minus one.
With no match at all, no relationship, you get a value
somewhere around zero. But consider this: if you have a whole lot of golf balls
bouncing around together on a concrete floor, quite randomly, some of them will
move together, just by chance.
There’s no cause, nothing in it at all, just a chance
matching up. And random variables can match up in the same way, just by chance.
And sometimes, that matchingup may have no meaning at all. This is why we have
tests of significance. We calculate the probability of getting a given
correlation by chance, and we only accept the fairly improbable values, the
ones that are unlikely to be caused by mere chance.
We aren’t on safe ground yet, because all sorts of wildly improbable
things do happen by chance. Winning the lottery is improbable, though
the lotteries people won't like me saying that. But though it's highly
improbable, it happens every day, to somebody. With enough tries, even the most
improbable things happen.
So here's why you should look around for some plausible link
between the variables, some reason why one of the variables might cause
the other. But even then, the lack of a link proves very little either way.
There may be an independent linking variable.
Suppose smoking was a habit which most beer drinkers had,
suppose most beer drinkers ate beer nuts, and just suppose that some beer nuts
were infected with a fungus which produces aflatoxins that cause slow cancers
which can, some years later, cause secondary lung cancers.
In this case, we'd get a correlation between smoking and lung
cancer which still didn't mean smoking actually caused lung cancer. And
that's the sort of grim hope which keeps those drug pushers, the tobacco czars
going, anyhow. It also keeps the smokers puffing away at their cancer sticks.
It shouldn't, of course, for people have thrown huge stacks
of variables into computers before this. The only answer which keeps
coming out is a direct and incontrovertible link between smoking and cancer.
The logic is there, when you consider the cigarette smoke, and how the amount
of smoking correlates with the incidence of cancer. It's an open and shut case.
I'm convinced, and I hope you are too. Still, just to
tantalise the smokers, I'd like to tell you about some of the improbable things
I got out of the computer in the 1980s. These aren't really what you might call
damned lies, and they are only marginally describable as statistics, but they
show you what can happen if you let the computer out for a run without a tight
lead.
Now anybody who's been around statistics for any time at all
knows the folklore of the trade, the old faithful standbys, like the price of
rum in Havana being highly correlated with the salaries of Presbyterian ministers
in Massachusetts, and the Dutch (or sometimes it's Danish) family size which
correlates very well with the number of storks' nests on the roof.
More kids in the house, more storks on the roof. Funny, isn't
it? Not really. We just haven't sorted through all of the factors yet. The
Presbyterian rum example is the result of correlating two variables which have
increased with inflation over many years.
You could probably do the
same with the cost of meat and the average salary of a vegetarian, but that
wouldn't prove anything much either. In the case of the storks on the roof,
large families have larger houses, and larger houses in cold climates usually
have more chimneys, and chimneys are what storks nest on. So naturally enough,
larger families have more storks on the roof. With this information, the
observed effect is easy to explain, isn't it?
There are others, though, where the explanation is less easy.
Did you know, for example, that Hungarian coal gas production correlates very
highly with Albanian phosphate usage? Or that South African paperboard
production matches the value of Chilean exports, almost exactly?
Or did you know the number of iron ingots shipped annually
from Pennsylvania to California between 1900 and 1970 correlates almost
perfectly with the number of registered prostitutes in Buenos Aires in the same
period? No, I thought you mightn't.
These examples are probably just a few more cases of two
items with similar natural growth, linked in some way to the world economy, or
else they must be simple coincidences. There are some cases, though, where, no
matter how you try to explain it, there doesn't seem to be any conceivable
causal link. Not a direct one, anyhow.
There might be indirect causes linking two things, like my
hypothetical beer nuts. These cases are worth exploring, if only as sources of
ideas for further investigation, or as cures for insomnia. It beats the hell
out of calculating the cube root of 17 to three decimal places in the wee small
hours, my own favourite gotosleep trick.
Now let's see if I can frighten you off listening to the
radio, that insomniac's standby. Many years ago, in a nowforgotten source, I
read that there was a very high correlation between the number of wireless
receiver licences in Britain, and the number of admissions to British mental
institutions.
At the time, I noted this with a wan smile, and turned to the
next taxing calculation exercise, for in those faroff days, all correlation
coefficients had to be laboriously handcalculated. It really was a long
time ago when I read about this effect.
It struck me, just recently, that radio stations pump a lot
of energy into the atmosphere. In America, the average fiveyearold lives in a
house which, over the child's life to the age of five, has received enough
radio energy to lift the family car a kilometre into the air. That's a lot
of energy.
Suppose, just suppose, that all this radiation caused some
kind of brain damage in some people. Not all of them necessarily, just a
susceptible few. Then, as you get more licences for wireless receivers in
Britain, so the BBC builds more transmitters and more powerful transmitters,
and more people will be affected. And so it is my sad duty to ask you all: are
the electronic media really out to rot your brains? Will cable TV save us
all?
Presented in this form, it's a contrived and, I hope,
unconvincing argument. Aside from anything else, the radiation is the wrong
wavelength and cannot change any cells. My purpose in citing these examples is
to show you how statistics can be misused to spread alarm and despondency. But
why bother?
Well, just a few years ago, problems like this were
rare. As I mentioned, calculating just one correlation coefficient was hard
yakka in the bad old days. Calculating the several hundred correlation
coefficients you would need to get one really improbable lulu was
virtually impossible, so fear and alarm seldom arose.
That was before the day of the personal computer and the hand
calculator. Now you can churn out the correlation coefficients faster than you
can cram the figures in, with absolutely no cerebral process
being involved.
As never before, we need to be warned to approach statistics
with, not a grain, but a shovelful, of salt. The statistic which can be
generated without cerebration is likely also to be considered without
cerebration. Which brings me, slowly but inexorably to the strange matter of
the podiatrists, the public telephones, and the births.
Seated one night at the keyboard, I was weary and ill at
ease. I had lost one of those essential connectors which link the parts of
one's computer. Then I found the lost cord, connected up my computer, and fed
it a huge dose of random data.
I found twenty ridiculously and obviously unrelated things,
so there were one hundred and ninety correlation coefficients to sift through.
That seemed about right for what I was trying to do.
When I was done, I switched on the printer, and sat back to
wait for the computer to churn out the results of its labours. The first few
lines of printout gave me no comfort, then I got a good one, then nothing
again, then a real beauty, and so it went: here are my cunningly selected
results. I have simply used, for good reasons, the methods of the crooks and
conmen.
Tasmanian
birth rate

SA
public phones

NSW
podiatrist registrations


Tasmanian
birth rate

1

+0.94

0.96

SA
public phones

+0.94

1

0.98

NSW
podiatrist registrations

0.96

0.98

1

Well of course the podiatrists and phones part is easy.
Quite clearly, New South Wales podiatrists are moving to South Australia and
metamorphosing into public phone boxes. Or maybe they're going to Tasmania to
have their babies, or maybe Tasmanians can only fall pregnant in South
Australian public phone booths.
Or maybe codswallop grows in computers which are treated unkindly.
Figures can't lie, but liars can figure. I would trust statistics any day, so
long as I can find out where they came from, and I'd even trust
statisticians, so long as I knew they knew their own limitations.
Most of the professional ones do know their limitations: it's the
amateurs who are dangerous.
I'd even use statistics to choose the safest hospital to go
to, if I had to go. But I'd still rather not go to hospital in the first place.
After all, statistics show clearly that more people die in the average
hospital than in the average home.
The original version is here: