>  The  "professional"  definition  looks  like  a "marketing numbers"
> game...

It  may  well, but it's the only definition. Your metric is valid, but
it needs a different term.

The  false  positive  rate  is  absolutely,  undeniably  a "quality or
accuracy  indicator,"  and  has  complete  acceptance  in research and
analysis. One thing that may be frustrating you is that it's used when
medical  research  frames  a  erroneously  "positive"  test  result as
relatively  harmless, but an erroneous "negative" as bad--such as your
"positive"  test  for  monkeypox,  when a person doesn't have it after
further  scrutiny, no big deal--in some senses, the opposite from spam
analysis.    Anti-spam    vendors   and   pundits   have   long   been
mixing-and-matching willy-nilly from concepts like absence/presence of
spamminess, positive/negative probability of non-spamminess, etc....in
other  words,  we've  probably  all  been  misusing, or inconsistently
using,  scientific  lingo from the get-go, and if "positive" itself is
ambiguous,  certainly  so will be "false positive" and "false positive
rate" be confusing.

Back on the scientific note, your example of a 1/99,991 false positive
rate being linked to 10 test positives in a monkeypox test would never
pass  muster  in  medicine  if  the  data  were  not  at or above some
previously  agreed-upon  level of statistical significance (p1 < p2 =<
.05),  and  it's  unlikely  that  a  disease  that's  only  present in
9/100,000  of the population would be able to be tested using a sample
size of only 100,000.

The  anti-spam  world  has  not,  however,  unified  on  the  specific
parameters  necessary to publish results using well-defined statistics
terms.  So  you're  correct  in  seeking  the  raw test results behind
marketing  claims;  in  general,  anyone  who  won't  provide raw test
results  on  request  is  not trustworthy, and many such offenders are
comfortably  within  the  medicine and scientific establishments, both
industrial and academic. Then, in the non-peer-reviewed world of which
anti-spam  is  a  part,  you'd  probably  be  astounded  at  how  many
over-the-counter  testing kits, local-news surveys, weight-loss pills,
and  the  like,  generate  false positives as often as true positives.
Perhaps we should stick, rather, with neologisms such as "Ham" (though
these  are  derided  by many who think we already have our jargon down
and should stick with it, a Catch-22).

> The  totally  separate "false negative" evaluation is: of the 99,990
> decisions "this person doesn't have monkey pox", how many really did
> have monkey pox (false negative)?

FTR,  the false negative rate is the percentage of those who are truly
negative who tested as positive. That is, if the test missed 3 people,
the false negative rate would be 3/12.

-Sandy


------------------------------------
Sanford Whiteman, Chief Technologist
Broadleaf Systems, a division of
Cypress Integrated Systems, Inc.
e-mail: [EMAIL PROTECTED]
------------------------------------


To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html
List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/
Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/

Reply via email to