Pablo, you are indeed correct (if you are talking about a clinical trial least, then you have a potential subgroup by treatment interaction), please see:
Subgroup analysis and other (mis)uses of baseline data in clinical trials.
The Lancet, Volume 355, Issue 9209, Pages 1064-1069
S. Assmann, S. Pocock, L. Enos, L. Kasten
Brookes ST, Whitley E, Peters TJ, Mulheran PA, Egger M, Davey Smith G. Subgroup analyses in randomised controlled trials: quantifying the risks of false-positives and false-negatives. Health Technol Assess 2001;5(33).
These are both available online. Beware the HTA monograph is over 60 pages long in case you hit print as second nature like me.
HTH,
Regards,
Graeme.
Graeme MacLennan
Statistician
Health Services Research Unit
Health Sciences Building
University of Aberdeen
Foresterhill
Aberdeen AB25 2ZD
Tel: +44 (0)1224 553487
Fax +44 (0)1224 554580
Email:
[hidden email]
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of Pablo A. Mora
Sent: 10 May 2007 15:27
To:
[hidden email]
Subject: Reference re: Interaction vs different regression models
I have the following problem, I tested a hypothesized triple interaction
term by using a cox regression model. One of the variables in the
interaction term is sex (female vs males). A reviewer criticize the
interaction approach and requested that I conduct different analyses for
males and females. I remember reading a while ago that if one has
hypotheses involving moderating effects, using interactions is the right way
to proceed. Running separate analyses, in this case by sex, could show
significant effects that are non-significant when tested through
interactions. Does anyone have a reference dealing with this particular
issue,that is, testing moderating effects with interactions vs. separate
models?
Thanks,
Pablo
Locked
