proc glimmixmethod=rspl data=a;
class community;
model disease(event='1')=x1-x17 w1/sdist=binary link=logit ddfm=bw;
random int w1 /subject= community;
nloptions tech=nrridg;
run;
proc nlmixed;
parms proc nlmixeddata=a;
parms b0=-2.4527b1=-0.3213b2=0.146b3=-0.2197 b4=0.05188b5=-0.03308b6=-0.1005b7=-0.3208
b8=0.1143b9=-0.046b10=-0.1553b11=-0.06697b12=0.1056b13=0.5637b14=0.1024b15=-0.1715
b16=0.02801b17=-0.1264b18=-0.06664b19=-0.03676b20=0.03011b21=-0.06736b22=0.009871
b23=0.04892b24=-0.1606b25=0.7477v=1;
logodds=b0+b1*x1+b2*x2+b3*c3+b4*c4+b5*x5+b6*x6_1+b7*x6_2+b8*x6_3+b9*x7_1+
b10*x7_2+b11*x7_3+b12*x7_4+b13*x8+b14*x9+b15*x10+b16*x11+b17*x12+b18*x13
+b19*x14+b20*x15_1+b21*x15_2+b22*x15_3+b23*x16+b24*x17+b25*w1+u;
odds=exp(logodds);
p=odds/(1+odds);
model a4~binary(p);
random u~normal(0,v) subject= community;
run;
estimate'var(u)' v;
run;
日前，在一篇硕士论文上面看到两水平logistic模型sas程序的一段编码，该模型将W1作为水平2解释变量，x1-x17作为水平1解释变量，拿出来与大家分享。
但是也有一点疑惑，sas程序是如何识别将W1 作为水平2解释变量的，还望有高人指点......