From Eric Kvaalen
Your article on sugar and alcoholism states that a simple test can identify
85 per cent of alcoholics
(This Week, 30 May, p 25). I have a test that can
identify 100 per cent of alcoholics: if they are male or female, then they are
alcoholics, otherwise they are not. (Unfortunately my test at present identifies
a lot of people as being alcoholics when they’re not.)
Later in the article, it is stated that the test correctly classified 66 out
of 78 volunteers, one-third of whom were alcoholics. This comes out to 85 per
cent. On this group, my test would have a success rate of only 33 per cent, but
on a group containing only alcoholics, it would do much better than the sugar
test.
The point is, the quality of a test is a function of both its false negative
rate (alcoholics identified as non-alcoholics) and its false positive rate
(non-alcoholics identified as alcoholics). The article doesn’t give either.
Qiryat Bialik, Israel
