Date: Mon, 22 Jun 2009 09:27:36 -0500
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: "Non-est" issue in the Proc Mixed Procedure
In-Reply-To: <722f50c2-6678-4b2d-a8ab-52042ffa6e20@w9g2000pro.googlegroups.com>
Content-Type: text/plain; charset="US-ASCII"
Stanley,
I usually find the most helpful way to write ESTIMATE statements, is to
first note the order of the variables on the class statement and the
number of levels for each,
CLASS subjid time paramn trt;
to write out a path with all combinations:
/*
time 0 1 2 3 4
5
paramn 1 2 1 2 1 2 1 2 1 2
1 2
trt 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1
2 1 2 1 2
*/
< note that the order on the CLASS statement is crucial to the next steps;
different orders produce different coefficients >
It is then rather straightforward to construct the estimates from these
placeholders:
ESTIMATE 'trt=1 at week3, paramn=100'
intercept 1
paramn 1 0
time 0 0 0 1 0 0
trt 1 0
time*trt 0 0 0 0 0 0 1 0 0 0 0 0
time*paramn 0 0 0 0 0 0 1 0 0 0 0 0
trt*paramn 1 0 0 0
time*trt*paramn 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 ;
ESTIMATE 'trt=2 at week3, paramn=100'
intercept 1
paramn 1 0
time 0 0 0 1 0 0
trt 0 1
time*trt 0 0 0 0 0 0 0 1 0 0 0 0
time*paramn 0 0 0 0 0 0 1 0 0 0 0 0
trt*paramn 0 1 0 0
time*trt*paramn 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 ;
and subtract the two statements to find the contrast "estimate" of
interest:
ESTIMATE 'trt=1 vs trt=2 at week3, paramn=100'
trt 1 -1
time*trt 0 0 0 0 0 0 1 -1 0 0 0 0
trt*paramn 1 -1 0 0
time*trt*paramn 0 0 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 /cl
;
run;
Robin High
UNMC
Stanley Wei <medpower@GMAIL.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
06/22/2009 04:05 AM
Please respond to
Stanley Wei <medpower@GMAIL.COM>
To
SAS-L@LISTSERV.UGA.EDU
cc
Subject
"Non-est" issue in the Proc Mixed Procedure
Dear Friends,
We encountered a problem when we tried to get the estimates with the
estimate statement in the PROC MIXED procedure, SAS always indicated
"Non-est" in the output window and we still do not figure out what's
the reason.
Sample data:
SUBJID TRT TIME PARAMN AVAL
329 1 0 250 9.200
329 1 1 250 9.815
329 1 2 250 9.800
329 1 3 250 10.000
329 1 4 250 9.900
329 1 5 250 9.900
332 2 0 100 228.000
332 2 1 100 244.000
332 2 2 100 231.000
332 2 3 100 202.000
332 2 4 100 181.000
332 2 5 100 181.000
332 2 0 250 7.800
332 2 1 250 8.416
332 2 2 250 7.500
332 2 3 250 7.300
332 2 4 250 6.900
332 2 5 250 6.900
369 2 0 100 154.000
369 2 1 100 174.000
369 2 2 100 127.000
369 2 3 100 137.000
369 2 4 100 148.000
369 2 5 100 170.000
369 2 0 250 8.800
369 2 1 250 7.772
369 2 2 250 7.700
369 2 3 250 7.200
369 2 4 250 7.400
369 2 5 250 7.800
373 1 0 100 270.000
373 1 1 100 235.000
373 1 2 100 261.000
373 1 3 100 232.000
373 1 4 100 173.000
373 1 5 100 160.000
... ...
Time variable has 6 levels, week 0, 1, 2, 3, 4, 5
Paramn has 2 levels 100, 250, indicate two different parameters
Trt has 2 levels treatment 1 and treatment 2
PROC MIXED DATA=test order=internal;
CLASS subjid time paramn trt;
MODEL aval = trt paramn time time*trt time*paramn trt*paramn
time*trt*paramn;
REPEATED paramn*time / SUBJECT=subjid TYPE=un;
ESTIMATE 'trt1 vs trt2 at week3 on A1C'paramn 1 0 time 0 1 0 0
0
0 trt 1 -1
time*trt 0 0 1 -1 0 0 0 0 0 0 0 0
time*paramn 0 0 1 0 0 0 0 0 0 0 0 0 intercept 1
trt*paramn 1 0 -1 0
time*trt*paramn 0 0 1 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0
0/cl;
ESTIMATE 'trt1 vs trt2 at week6 on A1C' trt 1 -1 paramn 1 0
time 0
0 1 0 0 0
time*trt 0 0 0 0 1 -1 0 0 0 0 0 0
time*paramn 0 0 0 0 1 0 0 0 0 0 0 0 intercept 1
trt*paramn 1 0 -1 0
time*trt*paramn 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 /
cl;
ESTIMATE 'trt1 vs trt2 at week12 on A1C' trt 1 -1 paramn 1 0
time
0 0 0 1 0 0
time*trt 0 0 0 0 0 0 1 -1 0 0 0 0
time*paramn 0 0 0 0 0 0 1 0 0 0 0 0 intercept 1
trt*paramn 1 0 -1 0
time*trt*paramn 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 /
cl;
ESTIMATE 'trt1 vs trt2 at week18 on A1C' trt 1 -1 paramn 1 0
time
0 0 0 0 1 0
time*trt 0 0 0 0 0 0 0 0 1 -1 0 0
time*paramn 0 0 0 0 0 0 0 0 1 0 0 0 intercept 1
trt*paramn 1 0 -1 0
time*trt*paramn 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 -1 0
0 /cl;
ESTIMATE 'trt1 vs trt2 at week24 on A1C' trt 1 -1 paramn 1 0
time
0 0 0 0 0 1
time*trt 0 0 0 0 0 0 0 0 0 0 1 -1
time*paramn 0 0 0 0 0 0 0 0 0 0 1 0 intercept 1
trt*paramn 1 0 -1 0
time*trt*paramn 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
-1 /cl;
ODS OUTPUT Estimates=outm1 covparms=rmse;
RUN;
Does anybody know what's the reason of "Non-est" issue with the codes
above? Is that caused by the data issues or some additional options to
be used or something else wrong with the codes?
Many thanks in advance for your help!
Stanley
