Date:         Mon, 17 Nov 2008 09:06:45 -0600
Reply-To:     Robin R High <rhigh@UNMC.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Robin R High <rhigh@UNMC.EDU>
Subject:      Re: Urgent help with power related for binomial proportion
Comments: To: Sophia Tong <sophiDT@HOTMAIL.COM>
In-Reply-To:  <200811162315.mAGBmJKU003090@malibu.cc.uga.edu>
Content-Type: text/plain; charset="US-ASCII"

Sophia,

PROC POWER won't let you solve for "delta" directly with the known inputs,
though with a bit of trial and repeated runs you can find a delta that
matches the input data you have, that is run the following code by
entering what you do know and modify the second value in the
GROUPPROPORTIONS option - xxxx -- until the value for p2 gives you .8
power:


PROC POWER;
TWOSAMPLEFREQ
   TEST = pchi /*other options: fisher lrchi pchi*/
   SIDES = 2
   ALPHA = .025
   GROUPPROPORTIONS = (p1 xxx)
   GROUPNS =  (n1 n2)
   POWER = .
;
RUN;

suppose p1=.2, n1=85, n2=90, with the values for alpha and number of
sides, then

PROC POWER;
TWOSAMPLEFREQ
   TEST = pchi /*fisher lrchi pchi*/
   SIDES = 2
   ALPHA = .025
   GROUPPROPORTIONS = (.2 .4135)
   GROUPNS =  (85 90)
   POWER = .
;
RUN;

Power here is close to .8, so delta would be .4135-.2 = .2135

Robin High
UNMC





Sophia Tong <sophiDT@HOTMAIL.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
11/16/2008 05:17 PM
Please respond to
Sophia Tong <sophiDT@HOTMAIL.COM>


To
SAS-L@LISTSERV.UGA.EDU
cc

Subject
Urgent help with power related for binomial proportion






Dear SAS -L listers,

I am trying to calculate minimum detectable difference in a proposed study
to cpmare with historical data. The sample size, proportion of historical
data are known (n1 and P1 are know known). Power and Alpha are arbitarily
decided as 0.8 and 0.025, sample size was also proposed for different
numbers ( so, n2, beta and alpha are all known). I need find delta.

I was trying to derive delta from sample size formular in a textbook and
give it up after finding it's beyond my knowledge.

Does anyone know how to do it in SAS, will PROC POWER do it? Or any
formular people have already figured it out?
My deadline is tomorow, you help is greatly appreciated.

Thanks

Sophia