Date:         Mon, 1 Dec 2008 15:00:53 -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: test for difference
Comments: To: Ben <benliu1227@GMAIL.COM>
In-Reply-To:  <f1de55c1-0f6a-45ea-b90a-d35d570bedb5@j39g2000yqn.googlegroups.com>
Content-Type: text/plain; charset="US-ASCII"

Ben,

One way, as outlined in section 9.1 of Fleiss, et. al., "Statistical
Methods for Rates and Proportions" 3rd. ed., is to run a chi-square test
with PROC FREQ. Since you have several "small" expected counts, include
the exact option.

DATA rsp;
INPUT cond death tot;
y='dead_'; count=death; ratio=death/tot; output;
y='alive'; count=tot-death; ratio=(tot-death)/tot; output;
cards;
1 6 37
2 3 30
3 1  9
4 2 26
;

proc freq data=rsp ORDER=data;
WEIGHT count ;
TABLE cond * y / NOCOL NOPERCENT expected chisq ;
EXACT chisq ;
RUN;


Robin High
UNMC




Ben <benliu1227@GMAIL.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
12/01/2008 12:54 PM
Please respond to
Ben <benliu1227@GMAIL.COM>


To
SAS-L@LISTSERV.UGA.EDU
cc

Subject
test for difference






i want to comapre the death ratio among 4 different groups which has
different health conditions,
so my data set looks like this,

type of condition           death ratio
1                                 6 out of 37 (16.2%)
2                                 3 out of 30 (10%)
3                                 1 out of 9   (11.1%)
4                                  2 out of 26 (7.7%)

so now i wanna do a stats test to see if the death ratios are the same
for the 4 types of condition,
but since there is only one observation for each type, i'm not sure
which analysis i should use.
please help, many thanks!