Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: sample size estimation

From   "Seed, Paul" <>
To   "" <>
Subject   st: RE: sample size estimation
Date   Wed, 16 May 2012 10:13:07 +0100

Dear Ofran, 

The sample size depends on what the 3 groups are, and what the 
planned method of analysis is.  

1) Are you talking about 3 treatment groups , or 3 patient groups?
I assume treatment groups. If they are patient groups, 
you are looking at subgroup analysis.

2) Are the 3 treatments 
	i) unordered (A, B, C)
	ii) partly ordered (control, A , B)
	iii) ordered (control, low dose, high dose)
Nick has addressed the case for design (i).
Design (ii) should be treated in the same way; however, there is 
a worrying tendency to carry out only 
the tests Control-A, Control-B, and not test A-B.
Authors then assume that if A is significantly different from Control
and B is not that A is different from B.  This error is very common.

Type (iii) opens up other possibilities - tests for trend, for deviation from trend, 

In all cases, your primary analysis should be specified in the protocol, and the 
sample size calculation adapted.


On Tue, May 15, 2012 at 11:08 AM, ofran almossawi
<> wrote:
> Hello
> Would anyone know how to do a sample size estimation to compare the
> effect of a treatment on three groups. I can only find information on
> how to compare two groups.
> Regards
> Ofran
> *

Paul T Seed, Senior Lecturer in Medical Statistics, 
Division of Women's Health, King's College London
Women's Health Academic Centre, KHP
(+44) (0) 20 7188 3642.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index