Unlike proc logistic, proc mixed uses an over-parameterized model without much flexibility on adjusting for design matrix. But you are still able to mange the parameterization through ORDER= option in PROC MIXED statement. In your situation, you can do things like:
- /*if you only ask for turning around one CLASS effect*/
- proc sort data =sashelp.class out =class; by descending age;
- run;
- proc mixed data =class order =data;
- class sex age;
- model weight =sex age/solution;
- run;
- /*for more than 1 effect, the above may not work, thereby you assort to format*/
- /*assigning values manually may be tedious, you can instead use data set to set formats*/
- proc format;
- value agefmt 16 ='1:16' 15 ='2:15'14 ='3:14'13 ='4:13'12 ='5:12'11='6:11';
- value $sexfmt F ='1:F' 'M' ='0:M';
- run;
- proc mixed data =class order =formatted;
- class sex age;
- format sex $sexfmt. age ageFmt.;
- model weight =sex|age/solution;
- run;
复制代码You may think it is not worth to do the manipulation when there is only such a little convenience gained from changing the default level. That is what I am thinking at least.
JingJu