Date:         Thu, 16 Mar 2006 16:19:28 -0600
Reply-To:     Robin High <robinh@UNLSERVE.UNL.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Robin High <robinh@UNLSERVE.UNL.EDU>
Subject:      Re: Sample-size determination for inequality of two proportions
Comments: To: Duck-Hye Yang <dyang@CHAPINHALL.ORG>
In-Reply-To:  <s41973fc.022@eden.chapinhall.org>
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Thu, 16 Mar 2006, Duck-Hye Yang wrote:

> Hello,
> I am looking for a way to do sample-size determination for a test of
> significance between two proportions.
> Proc power does not appear to provide a tool for a significance test
> for two proportions.
>

Hi Duckhye,

PROC POWER has at least two ways to invoke the TWOSAMPLEFREQ statement:

PROC POWER;
TWOSAMPLEFREQ
   TEST = LRchi /* options include Pchi LRchi fisher */
   SIDES = 2
   ALPHA = .05
   REFPROPORTION = .4
   PROPORTIONdiff =.15
   NPERGROUP = 50
   POWER = .
;
RUN;

* or..... ;

PROC POWER;
TWOSAMPLEFREQ
   TEST = LRchi
   SIDES = 2
   ALPHA = .05
   GROUPPROPORTIONS = (.4 .55)
   NPERGROUP = 50
   POWER = .
;
RUN;

To compare the result of Power =0.325 from these two steps, note that
according to Chapter 6 of Cohen's 1988 ed. of "Applied Power analysis"

  h = 2*arcsin(sqrt(.55)) - 2*arcsin(sqrt(.4)) = .3015

looking at Table 6.3.5 on pg. 195 for n=50, under the column for h=.3,
the power listed is "32" which seems to match the result SAS prints.

Robin High
Univ. of Oregon