[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: sample size for multiple group study

From   "Lachenbruch, Peter" <[email protected]>
To   <[email protected]>
Subject   st: RE: sample size for multiple group study
Date   Wed, 26 Dec 2007 13:15:54 -0800

This sounds a lot like a Dunnett test that compares k (=3) treatments to
a control.  You can check works on multiple testing such as R. Miller
(an old one) or try googling Dunnett's test.


Peter A. Lachenbruch
Department of Public Health
Oregon State University
Corvallis, OR 97330
Phone: 541-737-3832
FAX: 541-737-4001

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Scott
Sent: Wednesday, December 26, 2007 12:49 PM
To: [email protected]
Subject: st: sample size for multiple group study


I have a question about estimating the sample size needed for a 4-arm  
(placebo and 3 comparators) clinical study.  I've been reading the  
Statalist archives, Stata's FAQs, and the sampsi command, but I'm
still a bit confused.  I'm running Stata 9.2 on a Mac.  Any help or  
suggestions would be very much appreciated.

I want to estimate the number of folks needed for a study in which  
there are 4 groups, one of which is placebo, and the other 3 are  
different drugs.  Ideally, I'd like to compare each group to each  
other, but the main goal is to compare each drug to the placebo.   
There will be one pre-intervention assessment, and one post-  
intervention assessment.  The outcome is pain intensity, measured on  
a 0-10 numeric rating scale.  The mean pre-intervention score for  
everyone is expected to be 6.0, with an SD = 2.0.  The mean post- 
intervention scores are expected to be 4.5 (SD 2.0) for people who  
receive placebo, and 2.5 (SD 1.5) in each of the other 3 groups.  I'd  
like the power to be 0.9 and alpha to be 0.05.

 From the Stata documentation, it looks like my question is best  
addressed by the section on clinical trials with repeated measures,  
using a post approach, but I'm unsure how to handle to extra groups,  
since the sampsi command assumes two groups, and calculating sample  
size by hand is giving me dramatically different (and much larger)  
responses--roughly 90 people per group to detect a 3-unit  
difference.  From a previous study, the correlation between pre- 
intervention and post-intervention scores was 0.49.

sampsi 4.5 2.5, sd1(2.0) sd2(1.5) method(post) pre(1) post(1) r01(0.49)

Estimated sample size for two samples with repeated measures
          alpha =   0.0500  (two-sided)
                                       power =   0.9000
                                          m1 =      4.5
                                          m2 =      2.5
                                         sd1 =        2
                                         sd2 =      1.5
                                       n2/n1 =     1.00
            number of follow-up measurements =        1
             number of baseline measurements =        1
    correlation between baseline & follow-up =    0.490

Method: POST
  relative efficiency =    1.000
     adjustment to sd =    1.000
         adjusted sd1 =    2.000
         adjusted sd2 =    1.500

  Estimated required sample sizes:
                   n1 =       17
                   n2 =       17

Thank you,

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index