[S] Government regulations -- not

Gunter, Bert (bert_gunter@merck.com)
Wed, 29 Apr 1998 08:33:58 -0400


To Alan's wise comments, let me only add:

The government (FDA, whatever) requires nothing in terms of officially
mandated statistical analysis procedures. What you see here is internally
agreed-upon standards of FDA STATISTICIANS, who, as Alan made clear, must
strive for consistency in analytical procedures to avoid the appearance of
arbitrariness and inconsistency in their evaluation process. To which all
should pronounce a hearty amen, I think.

The problem therefore lies not with policy but practice, and here the fault,
dear Caesar, lies not with the stars but ourselves. For what Bill Venables
has really said is that the practices considered appropriate by government
statisticians are bad (in his opinion -- I have no comment on that). This
means that the primary task is educating each other -- what you see here is
a failure of the statistical community (and I would say ASA, in particular)
to provide appropriate forums for these issues to be discussed and best
current practice to be defined. Of course, FDA statisticians might help by
providing some design and analysis guidelines (perhaps they do, though my
impression is that most of this stuff is worked out via industry conferences
and meetings with individual companies). But what is really called for is
ASA (or ISI or RSS) or an appropriate section thereof to play a proactive
role in bringing together interested parties to determine agreed-upon best
practices. Unfortunately, ASA is dominated by academic interests and seems
little concerned with practice, as the rejection of any attempt to establish
statistical certification a few years ago seemed to confirm (but I don't
want to open this can of worms again). So we drift . Compare this with
standards setting in the engineering community (SAE,ASME, IEEE) and medical
communities in which the professional societies are much involved.

Even scarier is the thought that the problem might be that the stuff that is
taught in biostatistics programs is bad and these programs are so entrenched
in "traditional" statistics that they cannot change. I leave it to those of
you who teach in such programs to comment on that, however.

DISCLAIMER: The views expressed here are my own and not those of Merck & Co.
or my colleagues at Merck.

Bert Gunter
Biometrics Research
Merck Research Labs
P.O. Box 2000
Rahway, NJ 07065-0900
732-594-7765

"The business of the statistician is to catalyze the
scientific learning process." George E. P. Box

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news