Date: Thu, 30 Sep 2010 13:43:24 -0500 "KRISHNAMURTHY, SUBASH [AG/1000]" "SAS(r) Discussion" "KRISHNAMURTHY, SUBASH [AG/1000]" Re: Sample Size and Power for Multinomial Distribution? To: Robin R High A text/plain; charset="utf-8"

Robin, Thanks so much for your response. I will try your suggestion. The sample size and power is in the experimental design type situation, mainly a Randomized Complete Block Design or a Group block design.

This definitely helps.

-----Original Message----- From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Robin R High Sent: Thursday, September 30, 2010 1:39 PM To: SAS-L@LISTSERV.UGA.EDU Subject: Re: Sample Size and Power for Multinomial Distribution?

You can approximate power by entering an exmplary dataset to GLIMMIX, with the counts and follow the approach to power calculation examples from Chapter 12 of SAS for Mixed modes, 2nd ed.

one can add to the difficulty by thinking in terms of odds ratios based on glogits, or modify to clogits, but that's beyond what is presented below

* 3 response levels with input of proportions only tested across 2 groups;

%LET rw_t=125; * equal group sample size (row total); %LET p1a=.25; %LET p1b=.2; * p1c by subtraction, i.e., all p > 0 and sum(p1 p2 p3)=1; %LET p2a=.40; %LET p2b=.25; * p2c by subtraction;

* compute row probabilities;

DATA probs; grp =1 ; p1=&p1a.; p2=&p1b. ; p3=1-(p1+p2); output; grp =2 ; p1=&p2a.; p2=&p2b. ; p3=1-(p1+p2); output; run;

proc print; run;

* exemplary column counts for given group sample size (round to integers);

DATA counts; SET probs; keep r grp nn; r=1; nn=ROUND(p1*&rw_t.) ; OUTPUT; r=2; nn=ROUND(p2*&rw_t.) ; OUTPUT; r=3; nn=ROUND(p3*&rw_t.) ; OUTPUT; run;

ods output table=tbl(where=(_type_='10') keep=grp n _type_); PROC TABULATE data=counts NOSEPS; class grp r; FREQ nn; table grp, (r all)*n=' '*f=4.0 r*rowpctN='Row %'*f=7.1 / rts=9; run;

----------------------------------------------------- | | | | r | | | | |-----------------------| | | r | | 1 | 2 | 3 | | |--------------| |-------+-------+-------| | | 1 | 2 | 3 |All | Row % | Row % | Row % | |-------+----+----+----+----+-------+-------+-------| |grp | | | | | | | | |1 | 31| 25| 69| 125| 24.8| 20.0| 55.2| |2 | 50| 31| 44| 125| 40.0| 24.8| 35.2| -----------------------------------------------------

* because of rounding, actual row percents slightly different than inputs ;

proc transpose data=tbl out=t_c(drop=_name_) prefix=_; var n; id grp; proc print data=t_c; run;

ODS output tests3=tst; ods listing close; PROC GLIMMIX DATA=counts; CLASS grp ; MODEL r = grp / DIST=multinomial link=glogit; FREQ nn; run; ods listing;

proc print data=tst; run;

DATA pwr; SET tst; set t_c; DROP probF Fcrit fvalue Noncen numdf dendf _1 _2 ; NonCen = NumDF*Fvalue; alpha = .05; Fcrit = FINV(1-alpha, NumDF, DenDF, 0); power = 1- PROBF(Fcrit, NumDF, DenDF, Noncen); totaln=_1 + _2; RUN;

proc print data=pwr NOobs; run;

* gives an approximation ;

Effect alpha power totaln

grp 0.05 0.83299 250

Robin High UNMC

From: "KRISHNAMURTHY, SUBASH [AG/1000]" <subash.krishnamurthy@MONSANTO.COM> To: SAS-L@LISTSERV.UGA.EDU Date: 09/30/2010 09:33 AM Subject: Sample Size and Power for Multinomial Distribution? Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>

Hi,

Is there a good example to estimate Power and Sample Size based on Multinomial distribution?

Thanks. -Subash

This e-mail message may contain privileged and/or confidential information, and is intended to be received only by persons entitled to receive such information. If you have received this e-mail in error, please notify the sender immediately. Please delete it and all attachments from any servers, hard drives or any other media. Other use of this e-mail by you is strictly prohibited.

All e-mails and attachments sent and received are subject to monitoring, reading and archival by Monsanto, including its subsidiaries. The recipient of this e-mail is solely responsible for checking for the presence of "Viruses" or other "Malware". Monsanto, along with its subsidiaries, accepts no liability for any damage caused by any such code transmitted by or accompanying this e-mail or any attachment.

The information contained in this email may be subject to the export control laws and regulations of the United States, potentially including but not limited to the Export Administration Regulations ("EAR") and sanctions regulations issued by the U.S. Department of Treasury, Office of Foreign Asset Controls (“OFAC”). As a recipient of this information you are obligated to comply with all applicable U.S. export laws and regulations.

Back to: Top of message | Previous page | Main SAS-L page