st: RE: Sample size for a test of equivalence of proportions

["Philip Ryan" <philip.ryan@adelaide.edu.au>]

Angel

I confess I had quite forgotten I had written -equivsize-.  This is not too
surprising because it was a quick and dirty response to a question posted on
the List and not something destined for the SSC repository.  Reading my
original email, I see that I wrote: "There are no error traps and no help
file but comments embedded in the program tell the story. -equivsize-, in
*very* limited testing, appears to give the same answers as the commercial
program nQuery Advisor 4 (2000, Statistical Solutions, Cork)."  

I think all the above disclaimers provide me with sufficient cover for any
shortcomings in -equivsize-.

But in any case, I see that Philip Jones has recently released his program
-ssi- on the SSC and, although I have not used it, it looks to be much more
comprehensive than -equivsize-, is very (very) likely to be better tested
and should be more useful than -equivsize-.  Perhaps you should try -ssi- ?

Phil



Philip Ryan
Professor and Director
Data Management & Analysis Centre
School of Population Health & Clinical Practice
University of Adelaide


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Ángel Rodríguez
Laso
Sent: Wednesday, 26 May 2010 8:54 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: Sample size for a test of equivalence of proportions

Dear Statalisters,


I want to calculate the sample size for an experiment to compare the
refusal rate of two strategies to contact participants in a survey. In
the usual procedure, I expect a refusal rate of 17%. I would like to
detect a 10% increase in refusal with the alternative procedure, with
alpha 0.05 (one sided) and power 80%. I consider this a test of
equivalence between procedures.

After searching Statalist, I?ve come across two possibilities to
calculate sample sizes that yield different results: I would
appreciate your comments on possible misspecifications when entering
values for the programs and the reasons for the differences in the
results obtained.


1) The user-written program equivsize by Phil Ryan v0.02 2002-02-06
(http://www.stata.com/statalist/archive/2003-02/msg00204.html)

. equivsize .17 .17 .1 .05 .8

n (per group) = 175



2) The user-written artmenu program version 1.0.4 SB/PR 13jan2005:

. artbin, pr(.17 .27) ngroups(2) aratios(1 1) distant(0) alpha(0.05)
power(0.8) onesided(1) ni(1)

ART - ANALYSIS OF RESOURCES FOR TRIALS (version 1.0.0, 3 March 2004)
----------------------------------------------------------------------------
--
A sample size program by Abdel Babiker, Patrick Royston & Friederike
Barthel,
MRC Clinical Trials Unit, London NW1 2DA, UK.
----------------------------------------------------------------------------
--
Type of trial Noninferiority - binary outcome
Statistical test assumed Unconditional comparison of 2
binomial proportions
Number of groups 2
Allocation ratio Equal group sizes

Anticipated event probabilities 0.170, 0.270

Alpha 0.050 (one-sided)
Power (designed) 0.800

Total sample size (calculated) 451
Expected total number of events 100

That is, 225 individuals per group.


3) In addition, in a research methods text book (Argimón, Jiménez.
Métodos de investigación clínica y epidemiológica. Elsevier, Madrid,
2004) I?ve found the following formula to calculate sample sizes for
equivalence tests (no further reference for the formula is provided,
but maybe someone will identify its origin):

N(per group)= 2*P*(1-P)*(Zalpha+Zbeta)squared) / (difference of
interest)squared

Where P would be refusal rate in the usual procedure group (0.17)
Zalpha=1.645 (one-sided)
Zbeta=0.842 (corresponding to a 0.8 power)
Difference of interest=0.1

N(per group)= 2*0.17*0.83*(1.645+0.842)squared / 0.1*0.1 = 175


This matches perfectly the result of 1). Nevertheless, this option
does not allow unequal group sizes that could be of interest in this
experiment.



To complicate things further, the experiment will deal with clustered
samples. To correct sample size for clustering, would it be enough to
multiply the sample size obtained from these methods by the expected
DEFF?


Thank you very much for your time and interest.


Angel Rodriguez-Laso

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

No virus found in this incoming message.
Checked by AVG - www.avg.com 
Version: 9.0.819 / Virus Database: 271.1.1/2896 - Release Date: 05/26/10
03:56:00


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/