Date:         Wed, 17 Jan 2007 23:07:10 -0800
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: proportions test in SAS
In-Reply-To:  <cabe11870701170604s72b1aa89se6fa8d1418fc491a@mail.gmail.com>
Content-Type: text/plain; format=flowed

bgjiang@GMAIL.COM wrote back:
>
>On 1/17/07, LWn <lars.wahlgren.withoutthispart@stat.lu.se> wrote:
>>
>>Or perhaps this is what you're looking for:
>>
>>proc freq data=data ;
>>weight recip_no ;
>>table region*race
>>     / nopercent nocol chisquare ;
>>run ;
>>
>>HTH / LWn
>>
>>"baogong jiang" <bgjiang@GMAIL.COM> skrev i meddelandet
>>news:cabe11870701161316v6dcc790bha718f86f0bbba2b9@mail.gmail.com...
>> > Hello:
>> >
>> > Anyone knows how can I test if the proportion of the race is different
>>in
>> > all the regions?
>> > my data:
>> >
>> > DATA DATA;
>> > input region $ race $ recip_no;
>> > CARDS;
>> > region1 white 850
>> > region1 black 250
>> > region1 other 120
>> > region2 white 889
>> > region2 black 212
>> > region2 other 125
>> > region3 white 550
>> > region3 black 150
>> > region3 other 80
>> > region4 white 1520
>> > region4 black 1250
>> > region4 other 120
>> > ;run;
>> >
>> >
>> > thank you,
>> >
>> > baogong
>>
>
>
>
>--
>Baoogng Jiang
>Department of Agronomy
>Lousisana State University

>
>Thank you all for your suggestions. May be I did not describe my problem
>clearly, I am looking for an ANOVA like test to test the hypothesis: the
>proportions of different race (white, black, other) in different regions
>are
>the same.
>
>
>
>Can I use Logistic regression to test this hypothesis by considering race
>(white, black, other)  as multinomial outcome.
>
>
>
>*proc* *logistic* data=data rorder=data;
>
>class region ;
>
>model race=region;
>
>weight recip_no;
>
>*run*;
>
>
>
>or Can I use proc genmod to test the hypothesis like the following code:
>
>*proc* *genmod* data=data rorder=data;
>
>      freq recip_no;
>
>      class region;
>
>      model race = region / dist=multinomial
>
>                            link=cumlogit
>
>                            aggregate=region
>
>                            type3;
>
>*run*;
>
>
>
>Thank you very much,
>
>baogong
>

Perhaps you did not understand Lars' post completely.  His
code in PROC FREQ specifically looks at whether your RACE
variable has counts which are independent of your REGION
values.  That seems to be exactly what you are asking for.

Of course, if your data come from a sample survey, then PROC
FREQ - and PROC GENMOD and PROC LOGISTIC and ... - should
not be used.  Instead, you would need to use one of the SAS
survey analysis procs.

HTH,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
The MSN Entertainment Guide to Golden Globes is here.  Get all the scoop.
http://tv.msn.com/tv/globes2007/?icid=nctagline2