标题: 关于参照水平自定义的问题 SAS [打印本页] 作者: shiyiming 时间: 2012-4-24 00:18 标题: 关于参照水平自定义的问题 SAS 菜鸟求助个问题：
在研究无序多分类自变量X对二分类应变量的影响时，常采取LOGISTIC MODEL分析，捷报报告时，常DUMMY后，以X变量的某一水平如SAS默认以最高水平值）为REFERENCE LEVEL ,其他水平与其比较，计算 BETA ,EXP(BETA)等参数。 现在我想自定义参照组水平值，如以最低水平值为REFERENCE，试验如下：
logistic过程里PARAM语句可行；
proc logistic;
class x(param=ref ref=first);
model y =x/solution;
run;
但MIXED过程中 param语句无效。
proc mixed data=sasuser.lx1;
class no1 cishu (param=ref ref=first) ;
model ef=cishu/solution;
run;
请教：在MIXED模型中，是否也有同样类似语句。实在不想重新 recode新变量，多谢。作者: shiyiming 时间: 2012-4-24 06:11 标题: Re: 关于参照水平自定义的问题 SAS proc MIXED, similar to PROC GLM, uses GLM parameterization. That is, it won't take reference parameterization. But you can count on the future SAS version, i heard that that will be included in the next version(?).
without re-coding the variable, you can use the statement like estimate, contrast, lsmeans or means to get what you want. You may also know the POSTFIT function for sas starting from its verison of 9.2. Roughly, use STORE statement to record the data fitting informaiton and then you can get some results, such as estimates, lsmeans,etc. without re-fitting the model.
jingju作者: shiyiming 时间: 2012-4-24 14:13 标题: Re: 关于参照水平自定义的问题 SAS [quote="jingju11":1bzvbbfh]proc MIXED, similar to PROC GLM, uses GLM parameterization. That is, it won't take reference parameterization. But you can count on the future SAS version, i heard that that will be included in the next version(?).
without re-coding the variable, you can use the statement like estimate, contrast, lsmeans or means to get what you want. You may also know the POSTFIT function for sas starting from its verison of 9.2. Roughly, use STORE statement to record the data fitting informaiton and then you can get some results, such as estimates, lsmeans,etc. without re-fitting the model.
jingju[/quote:1bzvbbfh]
恩，谢谢京剧，或许该用contrast才能达到我预期目的了，不过那还不如直接 RECODE变量了，呵呵。