Dear All,
This mail is posted on behalf of Phil Woodward. All replies should
be sent to Phil at [log in to unmask]
John Parrott
____________________________________________________________________________
_________
I'm planning to organise a 1-day meeting, hopefully, in the early Summer of
2000 to discuss the role of p-values and significance testing in applied
statistics. This was prompted by John Nelder's paper in Part 2 of RSS
Series D this year. So far John Nelder, Stephen Senn and Andy Grieve have
agreed to talk and I would like a second person sympathetic to p-values,
preferably someone who works with engineers or in one of the "harder
sciences". The following provides a context for the meeting.
Do p-values and significance tests still have a role in applied statistics?
Should statisticians be more active in offering better alternatives to
scientists and engineers?
Andy Grieve
John Nelder
Stephen Senn
A.N. Other ( sympathetic to significance tests, ideally working with
engineers)
Although this issue has been discussed many times over the years it is still
true to say that:
p-values and significance tests dominate much statistical
teaching and software
most scientists and engineers equate statistical analysis
with significance testing
regulators and writers of guidelines promote significance
tests as the objective approach.
Clearly the subject is still of concern to many in the RSS, see Nelder (RSS
Series D, Vol. 48, Part 2, 1999) and the debate raised by Robert Matthews'
article in the Sunday Telegraph (RSS News, January 1999) for example.
Although it is true that many applied statisticians do regularly use
p-values and significance tests, it is believed there are relatively few who
see them any more than a potentially useful diagnostic measure rather than
the major purpose of the analysis. Although there are some application
areas, notably Phase III clinical trials, where statisticians focus on
significance testing, it appears that these methods are disproportionately
represented in our teaching, software and guidelines.
So, is it true that the majority of applied statisticians believe p-values
and significance tests, at best, add little additional information once
interval estimates have been calculated and, at worst, mislead scientists
and engineers into making the "prosecutor's fallacy", i.e. Pr(H | x) = Pr(x
| H)? Is it time to rethink the central position these methods hold in our
statistical teaching, software and guidelines?
The purpose of this meeting is to broaden the debate of this issue and
discuss what, if anything should be done.
If anyone would like to be considered for A.N.Other's position, please
contact me: [log in to unmask]
Phil Woodward
DDG (Biometrics)
B501/2.54 ipc:130 tel:48550
http://cr.sandwich.pfizer.com/biom/function/ddg/devres/welcome.htm
* Pfizer Central Research (c130), Ramsgate Rd, Sandwich, Kent, CT13 9NJ
* 01304 648550 (Fax: 653469)
* [log in to unmask]
LEGAL NOTICE
Unless expressly stated otherwise, this message is confidential and may be
privileged. It is intended for the addressee(s) only. Access to this e-mail
by anyone else is unauthorised. If you are not an addressee, any disclosure
or copying of the contents of this e-mail or any action taken (or not taken)
in reliance on it is unauthorised and may be unlawful. If you are not an
addressee, please inform the sender immediately. Pfizer Limited is
registered in England under No. 526209 with its registered office at
Ramsgate Road, Sandwich, Kent CT13 9NJ
John
John Parrott
DDG (Biometrics)
B501/2.48 ipc:130 tel:48641
http://cr.sandwich.pfizer.com/biom/function/ddg/index.htm
* Pfizer Central Research (c130), Ramsgate Rd, Sandwich, Kent, CT13 9NJ
* 01304 648641 (Fax: 653469)
* [log in to unmask]
LEGAL NOTICE
Unless expressly stated otherwise, this message is confidential and
may be privileged. It is intended for the addressee(s) only. Access to
this e-mail by anyone else is unauthorised. If you are not an
addressee, any disclosure or copying of the contents of this e-mail or
any action taken (or not taken) in reliance on it is unauthorised and
may be unlawful. If you are not an addressee, please inform the sender
immediately.
Pfizer Limited is registered in England under No. 526209 with its
registered office at Ramsgate Road, Sandwich, Kent CT13 9NJ
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|