"Gary Drescher" wrote in message
...
"Chip Jones" wrote in message
ink.net...

"Gary Drescher" wrote in message
...
"Chip Jones" wrote in message
k.net...
The method madated by the FAA ala a DOT 49 CFR Part 40 random urine
drug
test is called a GC/MS drug test. GC/MS drug testing has virtually
*no*
false positives...

You can read all about it he
http://www.aviationmedicine.com/drugtest.htm

Nothing at that web page states that the test has a low false-positive
rate.

The false-positive rate isn't even inherent in a test itself.
Specificity
is
an inherent property of a test--but even an excellent test with very
high
specificity can still have an arbitrarily high false-positive rate if
applied to a population in which positive examples are sufficiently
rare.
(The web page says nothing about the test's specificity either, by the
way.)

This is an issue that I can neither defend or attack, because I can't
quantify in my mind what an "arbitrarily high" false positive rate would
be
in a test group like commercial aviators, where positive examples are, I
agree, very rare.

Well, here are some illustrative numbers to help envision how an accurate
test could produce an arbitrarily high false-positive rate. Suppose the
test
has a specificity of 99% and also a sensitivity of 99%. Specificity refers
to the proportion of negative examples that correctly test negative;
sensitivity is the proportion of positive examples that correctly test
positive.

Now, suppose you apply this very accurate test to a population of one
million, among whom there are 1,000 positive examples. Among the 1,000
positive examples, about 990 will test positive, and about 10 will test
negative. Among the 999,000 negative examples, about 989,010 will test
negative, and about 9,900 will test positive.

Thus, among the 10,890 who test positive, 990 are actually positive
examples, and 9,900 are actually negative examples. Thus, the
false-positive
rate (the proportion of the positive test results that are false) is about
90.9%. Despite the use of an accurate test (99% sensitivity and
specificity), more than 90% of those who test positive will actually be
negative.

Here's one from NIDA, where the false positive rate in GC/MS drug
testing
for THC was 0.3 to 3.1%, and this was before any MRO action as per the
DOT
CFR.

Again, the false-positive rate *is not a function of just the accuracy of
the test*. A highly accurate test (high sensitivity and specificity) might
have an arbitrarily high or arbitrarily low false-positive rate, depending
on the proportion of actual positive and negative examples in the tested
population. Thus, a test's false-positive rate applied to one population
tells you nothing about the same test's false-positive rate applied to a
different population.

Thanks Gary, that's clearer to me now.

Chip, ZTL