Haven't seen any responses to this plea, and am therefore replying both
to "han" and to the edstat list. Comments and requests for
clarification are embedded in the original post, copied below.
On 20 Aug 2003, han wrote:
> Considering negative randomized controlled clinical trials, many of
> the therapies were labeled as no difference from control.
I understand "randomized controlled clinical trials" (I think), but am
not sure what you mean by "negative": is this another way of saying the
second part of the sentence, that the results did not permit rejection
of the null hypothesis, or do you intend something else altogether?
> How many factors should I take care in order not to miss an
> important therapy?
I do not understand the question. If you had asked, "how many trials
should I make in order to have probability > 0.80 (say) of detecting a
real effect of size {0.5 sigma, say}?", I would know what you were
asking and how to advise you to address the problem. But "factors"?
Usually by "factor" I would understand an independent variable thought
to have some interesting effect on a response variable of interest, and
an independent variable that occurred (in the experiment, at least) in
only a small number of discrete values. But you have not said how you
would identify (a) potentially useful factor(s), and so far as I can see
the question, under this interpretation of "factor", does not make a
great deal of sense.
> The type 1 error is fixed to 0.5.
I expect that what you meant was "The probability of a type 1 error is
fixed at 0.05." One might inquire what considerations led to adopting
this level of significance; you say nothing about the relative risks of
false negatives vs. false positives, nor the desired probability of a
type 2 error, so that one cannot tell whether 0.05 is reasonable.
(Of course, 0.5 is not, if you actually DID mean the probability of a
type 1 error.)
-----------------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
