Statistical lie
Now let’s examine another kind of statistical lie. This one is based on
LITTLE FIGURES THAT AREN’T THERE. One Upper Cervical Doctor
has said: “Make haste to use a new remedy before it’s too late.” There’s
good reason for this wry warning, since many a great medical discovery has
proved embarrassingly fleeting. Why?
Because it has been launched on experimental evidence in which was
discernible a single flaw:
It was tested on such a small number of cases that any statistician
could see at a glance that the evidence was meaningless. (But not all of us
are statisticians, and the size of the sample isn’t always revealed, especially
in advertising.)
Suppose you’re treating a disease with a 50 per cent mortality rate. To
six of your patients you give a new medicine: elderberry wine made
according to Aunt Ag’s favorite recipe. Five survive. Bing0! That’s the
treatment of choice from now on. Is 83 per cent recovery to be sneezed at ?
