btcm
我根据自己的实验建立了random coeff model，SAS程序如下：
data rc;
input treat$ time @@;
timec = time;
do i = 1 to 2;
input Y @@;
output;
end;
datalines;
t1 1 1.0000 1.0000
t1 2 0.6730 0.6484
t1 3 0.5734 0.5518
t1 4 0.5374 0.5100
t1 5 0.5064 0.4732
t1 6 0.4845 0.4554
t1 7 0.4233 0.3968
t2 1 1.0000 1.0000
t2 2 0.6202 0.6029
t2 3 0.4726 0.4505
t2 4 0.3892 0.3829
t2 5 0.3373 0.3309
t2 6 0.3175 0.3119
t2 7 0.2753 0.2697
t3 1 1.0000 1.0000
t3 2 0.5917 0.5793
t3 3 0.3578 0.3817
t3 4 0.2304 0.2659
t3 5 0.1573 0.1658
t3 6 0.1563 0.1344
t3 7 0.1391 0.1072
；
proc mixed data=rc;
class treat timec;
model Y = time / s;
random Int time timec / sub=treat s;
run;
请问各位大侠，我接下来想比较不同的treat（即t1、t2、t3的差异），该如何做？
先谢谢各位！
无痕
《SAS for Mixed Model 》里第八章，8.4.2
btcm
谢谢无痕 ！
下面是我修改了以后的程序：
data rc;
input treat$ time @@;
timec = time;
do i = 1 to 2;
input Y @@;
output;
end;
datalines;
t1 1 1.0000 1.0000
t1 2 0.6730 0.6484
t1 3 0.5734 0.5518
t1 4 0.5374 0.5100
t1 5 0.5064 0.4732
t1 6 0.4845 0.4554
t1 7 0.4233 0.3968
t2 1 1.0000 1.0000
t2 2 0.6202 0.6029
t2 3 0.4726 0.4505
t2 4 0.3892 0.3829
t2 5 0.3373 0.3309
t2 6 0.3175 0.3119
t2 7 0.2753 0.2697
t3 1 1.0000 1.0000
t3 2 0.5917 0.5793
t3 3 0.3578 0.3817
t3 4 0.2304 0.2659
t3 5 0.1573 0.1658
t3 6 0.1563 0.1344
t3 7 0.1391 0.1072
;
proc mixed data=rc;
class treat;
model Y = treat|time/ddfm=kr;
random Int time / type=un sub=treat s;
lsmeans treat / at time=1 diff;
lsmeans treat / at time=2 diff;
lsmeans treat / at time=3 diff;
lsmeans treat / at time=4 diff;
lsmeans treat / at time=5 diff;
lsmeans treat / at time=6 diff;
lsmeans treat / at time=7 diff;
ods output lsmeans=lsm diffs=diffs;
run;
运行后，LOG里面出现Convergence criteria met but final hessian is not positive definite.
在下不甚明了，请问各位大虾，可能是哪里出了问题？
